Untitled
Untitled
Untitled
Data-Driven Network
Security
Brij B. Gupta
National Institute of Technology, Kurukshetra, India
Copyright © 2022 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or
companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.
Library of Congress Cataloging-in-Publication Data
This book is published in the IGI Global book series Advances in Information Security, Privacy, and Ethics (AISPE) (ISSN:
1948-9730; eISSN: 1948-9749)
All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the
authors, but not necessarily of the publisher.
Mission
As digital technologies become more pervasive in everyday life and the Internet is utilized in ever in-
creasing ways by both private and public entities, concern over digital threats becomes more prevalent.
The Advances in Information Security, Privacy, & Ethics (AISPE) Book Series provides cutting-
edge research on the protection and misuse of information and technology across various industries and
settings. Comprised of scholarly research on topics such as identity management, cryptography, system
security, authentication, and data protection, this book series is ideal for reference by IT professionals,
academicians, and upper-level students.
Coverage
• Risk Management
IGI Global is currently accepting manuscripts
• Technoethics
for publication within this series. To submit a pro-
• Data Storage of Minors
posal for a volume in this series, please contact our
• Information Security Standards
Acquisition Editors at Acquisitions@igi-global.com
• Global Privacy Concerns
or visit: http://www.igi-global.com/publish/.
• CIA Triad of Information Security
• Security Information Management
• Telecommunications Regulations
• Privacy-Enhancing Technologies
• Cookies
The Advances in Information Security, Privacy, and Ethics (AISPE) Book Series (ISSN 1948-9730) is published by IGI Global, 701 E.
Chocolate Avenue, Hershey, PA 17033-1240, USA, www.igi-global.com. This series is composed of titles available for purchase individually;
each title is edited to be contextually exclusive from any other title within the series. For pricing and ordering information please visit http://
www.igi-global.com/book-series/advances-information-security-privacy-ethics/37157. Postmaster: Send all address changes to above address.
Copyright © 2022 IGI Global. All rights, including translation in other languages reserved by the publisher. No part of this series may be
reproduced or used in any form or by any means – graphics, electronic, or mechanical, including photocopying, recording, taping, or informa-
tion and retrieval systems – without written permission from the publisher, except for non commercial, educational use, including classroom
teaching purposes. The views expressed in this series are those of the authors, but not necessarily of IGI Global.
Titles in this Series
For a list of additional titles in this series, please visit: www.igi-global.com/book-series
Editorial Advisory Board
Ammar Almomani, Al-Balqa Applied University, Jordan
Nalin A. G. Arachchilage, University of New South Wales, Australia
V. C. Bhavsar, UNB, Canada
Mohd Anuaruddin Bin Ahmadon, Yamaguchi University, Japan
Arcangelo Castiglione, University of Salerno, Italy
Xiaojun Chang, Monash University, Clayton, Australia
Francisco José García Peñalvo, University of Salamanca, Spain
Deepak Gupta, LoginRadius Inc., Canada
Ching-Hsien Hsu, Asia University, Taiwan
Jin Li, Guangzhou University, China
Yining Liu, Guilin University of Electronic Technology, China
Ahmad Manasrah, Yarmouk University, Jordan
Gregorio Martinez Perez, University of Murcia, Spain
Melody Moh, San Jose State University, USA
Nadia Nedjah, State University of Rio de Janeiro, Brazil
Dragan Peraković, University of Zagreb, Croatia
T. Perumal, Universiti Putra Malaysia (UPM), Malaysia
Raffaele Pizzolante, University of Salerno, Italy
Konstantinos Psannis, University of Macedonia, Greece
Imran Razzak, Deakin University, Australia
Michael Sheng, Macquarie University, Australia
Shingo Yamaguchi, Yamaguchi University, Japan
Table of Contents
Preface.................................................................................................................................................. xvi
Acknowledgment................................................................................................................................. xxi
Chapter 1
Machine Learning for Malware Analysis: Methods, Challenges, and Future Directions........................ 1
Krishna Yadav, National Institute of Technology, Kurukshetra, India
Aarushi Sethi, National Institute of Technology, Kurukshetra, India
Mavneet Kaur, National Institute of Technology, Kurukshetra, India
Dragan Perakovic, University of Zagreb, Croatia
Chapter 2
Research Trends for Malware and Intrusion Detection on Network Systems: A Topic Modelling
Approach................................................................................................................................................ 19
Santosh Kumar Smmarwar, National Institute of Technology, Raipur, India
Govind P. Gupta, National Institute of Technology, Raipur, India
Sanjay Kumar, National Institute of Technology, Raipur, India
Chapter 3
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Data-Driven
Network Security................................................................................................................................... 41
Praneeth Gunti, National Institute of Technology, Kurukshetra, India
Brij B. Gupta, National Institute of Technology, Kurukshetra, India
Francisco José García Peñalvo, University of Salamanca, Spain
Chapter 4
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection....... 59
Kwok Tai Chui, Hong Kong Metropolitan University, Hong Kong
Patricia Ordóñez de Pablos, The University of Oviedo, Spain
Miltiadis D. Lytras, Deree College — The American College of Greece, Greece
Ryan Wen Liu, Wuhan University of Technology, China
Chien-wen Shen, National Central University, Taiwan
Chapter 5
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning
Techniques............................................................................................................................................. 74
Ángel Luis Perales Gómez, University of Murcia, Spain
Lorenzo Fernández Maimó, University of Murcia, Spain
Alberto Huertas Celdrán, University of Zurich, Switzerland
Félix Jesús García Clemente, University of Murcia, Spain
Chapter 6
Malicious Node Detection Using Convolution Technique: Authentication in Wireless Sensor
Networks (WSN).................................................................................................................................... 94
Priyanka Ahlawat, National Institute of Technology, Kurukshetra, India
Pranjil Singhal, National Institute of Technology, Kurukshetra, India
Khushi Goyal, National Institute of Technology, Kurukshetra, India
Kanak Yadav, National Institute of Technology, Kurukshetra, India
Rohit Bathla, National Institute of Technology, Kurukshetra, India
Chapter 7
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication.................. 112
Priyanka Ahlawat, National Institute of Technology, Kurukshetra, India
Kanishka Tyagi, National Institute of Technology, Kurukshetra, India
Chapter 8
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network................................ 127
Shingo Yamaguchi, Yamaguchi University, Japan
Brij Gupta, National Institute of Technology, Kurukshetra, India
Chapter 9
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)...... 148
Anish Khan, UIET, Kurukshetra University, Kurukshetra, India
Dragan Peraković, University of Zagreb, Croatia
Chapter 10
SecBrain: A Framework to Detect Cyberattacks Revealing Sensitive Data in Brain-Computer
Interfaces.............................................................................................................................................. 176
Enrique Tomás Martínez Beltrán, University of Murcia, Spain
Mario Quiles Pérez, University of Murcia, Spain
Sergio López Bernal, University of Murcia, Spain
Alberto Huertas Celdrán, University of Zürich, Switzerland
Gregorio Martínez Pérez, University of Murcia, Spain
Chapter 11
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications................ 199
Santosh Kumar Smmarwar, National Institute of Technology, Raipur, India
Govind P. Gupta, National Institute of Technology, Raipur, India
Sanjay Kumar, National Institute of Technology, Raipur, India
Chapter 12
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks
(WSN).................................................................................................................................................. 219
Priyanka Ahlawat, National Institute of Technology, Kurukshetra, India
Mukul Goyal, National Institute of Technology, Kurukshetra, India
Rishabh Sethi, National Institute of Technology, Kurukshetra, India
Nitish Gupta, National Institute of Technology, Kurukshetra, India
Chapter 13
Cybersecurity Risks Associated With Brain-Computer Interface Classifications............................... 236
Sergio López Bernal, University of Murcia, Spain
Alberto Huertas Celdrán, University of Zürich, Switzerland
Gregorio Martínez Pérez, University of Murcia, Spain
Index.................................................................................................................................................... 302
Detailed Table of Contents
Preface.................................................................................................................................................. xvi
Acknowledgment................................................................................................................................. xxi
Chapter 1
Machine Learning for Malware Analysis: Methods, Challenges, and Future Directions........................ 1
Krishna Yadav, National Institute of Technology, Kurukshetra, India
Aarushi Sethi, National Institute of Technology, Kurukshetra, India
Mavneet Kaur, National Institute of Technology, Kurukshetra, India
Dragan Perakovic, University of Zagreb, Croatia
Companies and organizations are collecting all sorts of data ranging from nominal feedback like customer
reviews to highly classified data like medical records. With data being such a critical aspect of most
of the operations around us, cybercriminals are looking for an opportunity to misuse this information.
One such device that cybercriminals use to further their malicious intent is malware. Over the years,
these cybercriminals have become immensely powerful using the knowledge of previous attacks. Hence,
malware analysis and methods to troubleshoot the problems arising due to malware attacks is the need
of the hour. Over time, different new approaches have been developed to defend malware. However, in
recent times, machine learning-based malware analysis has gained popularity. The capacity to detect
possible future malware by learning from existing malware patterns makes this method very popular.
In this chapter, the authors have introduced different malware and the machine learning-based approach
that has been developed in recent times to mitigate malware.
Chapter 2
Research Trends for Malware and Intrusion Detection on Network Systems: A Topic Modelling
Approach................................................................................................................................................ 19
Santosh Kumar Smmarwar, National Institute of Technology, Raipur, India
Govind P. Gupta, National Institute of Technology, Raipur, India
Sanjay Kumar, National Institute of Technology, Raipur, India
With more uses of internet-based services, the risk of cyberattacks is growing continuously. To analyze
these research trends for malware and intrusion detection, the authors applied the topic modeling approach
in the study by using the LDA (latent dirichlet allocation) and calculating the maximum and minimum
probability of the words, which appears in the large collection of text. The LDA technique is useful in
finding the hidden topics for further research in the areas of network and cybersecurity. In this chapter,
they collected the abstract of two thousand papers from the Scopus library from 2014 to 2021. These
collected papers are from reputed publications such as Elsevier, Springer, and IEEE Transactions. The
main aim of this study is to find research trends based on keywords that are untouched or on which less
research work has been done. To the best of the authors’ knowledge, this will be the first study done by
using the LDA technique for topic modeling in the areas of network security to demonstrate the research
gap and trends for malware and intrusion detection systems.
Chapter 3
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Data-Driven
Network Security................................................................................................................................... 41
Praneeth Gunti, National Institute of Technology, Kurukshetra, India
Brij B. Gupta, National Institute of Technology, Kurukshetra, India
Francisco José García Peñalvo, University of Salamanca, Spain
A never-ending fight is taking place among malware creators and security experts as the advances in
malware are daunting. The machine learning strategies are indeed the new mode of researching malware.
The purpose of this chapter is to explore machine learning methods for malware recognition and in general
deep learning methods. The chapter gives complete explanations of the techniques and resources used in
a standard machine learning process for detecting malware. It examines the study issues that are posed
by existing study methods and introduces the potential avenues of study in future. By administering a
study to the participants, scholars have a better knowledge of the malware detections. The authors start
by discussing simple dynamic modelling methods, their importance to the data analytics of malware,
and their implementations. They use open access resources such as virustotal.com that review sample
of dynamic analysis in reality.
Chapter 4
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection....... 59
Kwok Tai Chui, Hong Kong Metropolitan University, Hong Kong
Patricia Ordóñez de Pablos, The University of Oviedo, Spain
Miltiadis D. Lytras, Deree College — The American College of Greece, Greece
Ryan Wen Liu, Wuhan University of Technology, China
Chien-wen Shen, National Central University, Taiwan
Software has been the essential element to computers in today’s digital era. Unfortunately, it has experienced
challenges from various types of malware, which are designed for sabotage, criminal money-making,
and information theft. To protect the gadgets from malware, numerous malware detection algorithms
have been proposed. In the olden days there were shallow learning algorithms, and in recent years there
are deep learning algorithms. With the availability of big data for training of model and affordable and
high-performance computing services, deep learning has demonstrated its superiority in many smart city
applications, in terms of accuracy, error rate, etc. This chapter intends to conduct a systematic review
on the latest development of deep learning algorithms for malware detection. Some future research
directions are suggested for further exploration.
Chapter 5
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning
Techniques............................................................................................................................................. 74
Ángel Luis Perales Gómez, University of Murcia, Spain
Lorenzo Fernández Maimó, University of Murcia, Spain
Alberto Huertas Celdrán, University of Zurich, Switzerland
Félix Jesús García Clemente, University of Murcia, Spain
In the last decades, factories have suffered a significant change in automation, evolving from isolated
towards interconnected systems. However, the adoption of open standards and the opening to the
internet have caused an increment in the number of attacks. In addition, traditional intrusion detection
systems relying on a signature database, where malware patterns are stored, are failing due to the high
specialization of industrial cyberattacks. For this reason, the research community is moving towards the
anomaly detection paradigm. This paradigm is showing great results when it is implemented using machine
learning and deep learning techniques. This chapter surveys several incidents caused by cyberattacks
targeting industrial scenarios. Next, to understand the current status of anomaly detection solutions, it
analyses the current industrial datasets and anomaly detection systems in the industrial field. In addition,
the chapter shows an example of malware attacking a manufacturing plant, resulting in a safety threat.
Finally, cybersecurity and safety solutions are reviewed.
Chapter 6
Malicious Node Detection Using Convolution Technique: Authentication in Wireless Sensor
Networks (WSN).................................................................................................................................... 94
Priyanka Ahlawat, National Institute of Technology, Kurukshetra, India
Pranjil Singhal, National Institute of Technology, Kurukshetra, India
Khushi Goyal, National Institute of Technology, Kurukshetra, India
Kanak Yadav, National Institute of Technology, Kurukshetra, India
Rohit Bathla, National Institute of Technology, Kurukshetra, India
Considering the situation when the node which is entering in the network is a malicious node, it can check
and record the various operations of the network which could be responsible for some serious security
issues. In order to safeguard the network, the author propose an approach which is in view of the advanced
encoding strategy, termed as convolution coding approach. According to the requirement of network, the
initial bits are assigned, and by applying the convolution technique, the final code is generated for each
and every node. Considering the fact that it is a digital method, the codes can be represented by using
the binary number system. All the nodes that are a part of the network will be having their corresponding
final binary code, say C. The verification of each node is carried out by matching the generated code C
with the security code within a particular time period before transmitting the data. This process enables
the detection of malicious or attacker node. Furthermore, to enhance the versatility and execution, the
system is organized into clusters.
Chapter 7
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication.................. 112
Priyanka Ahlawat, National Institute of Technology, Kurukshetra, India
Kanishka Tyagi, National Institute of Technology, Kurukshetra, India
In the real scenario, there is a large multicast group where nodes leave and join frequently, and also the
number of nodes leaving and joining is also not proportionate. Hence, scalable rekeying process is an
important issue that needs to be concerned for the secured group communication for dynamic groups.
In basic rekeying scheme, which is based on the logical key hierarchy, the rekeying cost depends on the
logarithm of the size of group for a join or depart request by the user. However, the memory efficiency
of this group rekeying protocol (GREP) is a huge storage overhead over the system. The authors aim
to provide a survey of various group key management schemes and then propose an efficient scalable
solution based on linked LKH and the linked list data structure. Results have shown that the Linked LKH
algorithm has a very low effective cost for rekeying the LKH as compared to the basic LKH algorithm
(i.e., based on the number of new joined and departure requests).
Chapter 8
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network................................ 127
Shingo Yamaguchi, Yamaguchi University, Japan
Brij Gupta, National Institute of Technology, Kurukshetra, India
This chapter introduces a new kind of cybersecurity system named botnet defense system (BDS) that
defends an IoT system against malicious botnets. This chapter consists of two parts. The former part
describes the concept and design of the BDS. The concept is “fight fire with fire.” To realize the concept,
the BDS uses bot technology. The BDS builds a white-hat botnet on the IoT system by itself and uses it
to exterminate the malicious botnets. The white-hat botnet autonomously spreads over the IoT system
and thus drastically increases the defense ability. The latter part explains the strategy of the BDS. The
white-hat botnet is a so-called double-edged sword. It defends the IoT system against malicious botnet
but wastes the system’s resources. Therefore, the BDS should strategically use the white-hat botnet.
Some strategies have been proposed. Their characteristics are discussed through the simulation with
the agent-oriented petri nets.
Chapter 9
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)...... 148
Anish Khan, UIET, Kurukshetra University, Kurukshetra, India
Dragan Peraković, University of Zagreb, Croatia
The internet of things is a cutting-edge technology that is vulnerable to all sorts of fictitious solutions.
As a new phase of computing emerges in the digital world, it intends to produce a huge number of
smart gadgets that can host a wide range of applications and operations. IoT gadgets are a perfect target
for cyber assaults because of their wide dispersion, availability/accessibility, and top-notch computing
power. Furthermore, as numerous IoT devices gather and investigate private data, they become a gold
mine for hostile actors. Hence, the matter of fact is that security, particularly the potential to diagnose
compromised nodes, as well as the collection and preservation of testimony of an attack or illegal activity,
have become top priorities. This chapter delves into the timeline and the most challenging security and
privacy issues that exist in the present scenario. In addition to this, some open issues and future research
directions are also discussed.
Chapter 10
SecBrain: A Framework to Detect Cyberattacks Revealing Sensitive Data in Brain-Computer
Interfaces.............................................................................................................................................. 176
Enrique Tomás Martínez Beltrán, University of Murcia, Spain
Mario Quiles Pérez, University of Murcia, Spain
Sergio López Bernal, University of Murcia, Spain
Alberto Huertas Celdrán, University of Zürich, Switzerland
Gregorio Martínez Pérez, University of Murcia, Spain
In recent years, the growth of brain-computer interfaces (BCIs) has been remarkable in specific application
fields, such as the medical sector or the entertainment industry. Most of these fields use evoked potentials,
like P300, to obtain neural data able to handle prostheses or achieve greater immersion experience in
videogames. The natural use of BCI involves the management of sensitive users’ information as behaviors,
emotions, or thoughts. In this context, new security breaches in BCI are offering cybercriminals the
possibility of collecting sensitive data and affecting subjects’ physical integrity, which are critical issues.
For all these reasons, the fact of applying efficient cybersecurity mechanisms has become a main challenge.
To improve this challenge, this chapter proposes a framework able to detect cyberattacks affecting one of
the most typical scenarios of BCI, the generation of P300 through visual stimuli. A pool of experiments
demonstrates the performance of the proposed framework.
Chapter 11
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications................ 199
Santosh Kumar Smmarwar, National Institute of Technology, Raipur, India
Govind P. Gupta, National Institute of Technology, Raipur, India
Sanjay Kumar, National Institute of Technology, Raipur, India
Blockchain since 2009 has been gaining more popularity in various fields to use in numerous applications
to overcome the security issues such as privacy, transparency, and mutability of data in the process of
data sharing. Process of data sharing has many addressed and unaddressed challenges such as information
encryption and decryption, data authentication, storage security, latency time, transfer speed of data,
detecting malicious nodes, prevent the computer system from attacks, trust in the sharing process. In this
chapter, the authors have reviewed the data sharing paper based on blockchain technology and presented
the analysis of various techniques used in the information sharing process. The comprehensive analysis
is categorizing in the following areas like incentive mechanism-based work, IoT-based data sharing,
healthcare data sharing, and internet of vehicle data sharing using blockchain.
Chapter 12
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks
(WSN).................................................................................................................................................. 219
Priyanka Ahlawat, National Institute of Technology, Kurukshetra, India
Mukul Goyal, National Institute of Technology, Kurukshetra, India
Rishabh Sethi, National Institute of Technology, Kurukshetra, India
Nitish Gupta, National Institute of Technology, Kurukshetra, India
Node capture attack is one of the crucial attacks in wireless sensor networks (WSN) that seizes the
node physically and withdraws the confidential data from the node’s memory. The chapter exploits the
adversarial behavior during a node capture to build an attack model. The authors also propose a fruit
fly optimization algorithm (FFOA) that is a multi-objective optimization algorithm that consists of a
number of objectives for capturing a node in the network: maximum node contribution, maximum key
contributions are some examples of the same. The aim is to demolish the maximum part of the network
while minimizing the cost and maximizing attacking efficiency. Due to the multi-objective function, the
authors attain a maximum fraction of compromised traffic, lower attacking rounds, and lower energy
cost as contrasted with other node capture attack algorithms. They have developed an algorithm, which
is an enhanced version of FFOA and has even better efficiency than FFOA.
Chapter 13
Cybersecurity Risks Associated With Brain-Computer Interface Classifications............................... 236
Sergio López Bernal, University of Murcia, Spain
Alberto Huertas Celdrán, University of Zürich, Switzerland
Gregorio Martínez Pérez, University of Murcia, Spain
Brain-computer interfaces (BCIs) have experienced a considerable evolution in the last decade, expanding
from clinical scenarios to sectors such as entertainment or video games. Nevertheless, this popularization
makes them a target for cyberattacks like malware. Current literature lacks comprehensive works focusing
on cybersecurity applied to BCIs and, mainly, publications performing a rigorous analysis of the risks
and weaknesses that these interfaces present. If not studied properly, these potential vulnerabilities could
dramatically impact users’ data, service availability, and, most importantly, users’ safety. Because of
that, this work introduces an evaluation of the risk that each BCI classification already defined in the
literature presents to raise awareness between the readers of this chapter about the potential threat that
BCIs can generate in the next years if comprehensive measures, based on standard mechanisms, are not
adopted. Moreover, it seeks to alert academic and industrial stakeholders about the impact these risks
could have on future BCI hardware and software.
Index.................................................................................................................................................... 302
xvi
Preface
Every day approximately three-hundred thousand to four-hundred thousand new malware are registered,
many of them being adware and variants of previously known malware. Anti-virus companies and re-
searchers cannot deal with such a deluge of malware to analyze and build patches. The only way to scale
the efforts is to build algorithms to enable machines to analyze malware and classify and cluster them to
such a level of granularity that it will enable humans (or machines) to gain critical insights about them
and build solutions that are specific enough to detect and thwart existing malware and generic-enough
to thwart future variants.
Advances in Malware and Data-Driven Network Security comprehensively covers data-driven malware
security with an emphasis on using statistical, machine learning, and AI as well as the current trends in
ML/statistical approaches to detecting, clustering, and classification of cyber-threats. Providing informa-
tion on advances in malware and data-driven network security as well as future research directions, it
is ideal for graduate students, academicians, faculty members, scientists, software developers, security
analysts, computer engineers, programmers, IT specialists, and researchers who are seeking to learn and
carry out research in the area of malware and data-driven network security.
This book contains chapters dealing with different aspects of Malware analysis, Intrusion Detection on
Network System, Data-driven Network Security, Anti-Virus Vendors, Botnet Defense Systems, Malicious
Node Detection, Brain-Computer Interfaces Classifications, Cybersecurity Risks, Data-Driven Network
Security, Deep Learning Techniques, Blockchain Systems, Linked LKH Algorithms, WSN Security,
Machine Learning, Malicious Node Detection, Malware Algorithms, Malware Detection, SecBrain.
Specifically, this book contains discussion on the following topics:
• Machine Learning for Malware Analysis: Methods, Challenges, and Future Directions
• Research Trends for Malware and Intrusion Detection on Network System: A Topic Modelling
Approach
• Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Data-Driven
Network Security
• The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
• Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
• Malicious Node Detection using Convolution Technique: Authentication in Wireless Sensor
Network (WSN)
• Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
• Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
• A Survey on Emerging Security Issues, Challenges and Solutions for Internet of Things (IoTs)
Preface
ORGANIZATION OF BOOK
Chapter 1: Nowadays companies and organizations are collecting all sorts of data ranging from nomi-
nal feedback like customer reviews to highly classified data like medical records. With data being such
a critical aspect of most of the operations around us, cybercriminals are looking for an opportunity to
misuse this information. One such device that cybercriminals use to further their malicious intent is
malware. Over the years, these cybercriminals have become immensely powerful using the knowledge
of previous attacks. Hence malware analysis and methods to troubleshoot the problems arising due to
malware attacks is the need of the hour. Over time, different new approaches have been developed to
defend malware. However, in recent times, machine learning-based malware analysis has gained enough
popularity. The capacity to detect possible future malware by learning from existing malware patterns
makes this method very popular. In this chapter, we have introduced different malware and the machine
learning-based approach that has been developed in recent times to mitigate these malware.
Chapter 2: With more uses of internet-based services, the risk of cyberattacks is growing continuously,
to analyze these research trends for malware and intrusion detection, we applied the topic modeling
approach in our study by using the LDA (Latent Dirichlet Allocation) and calculate the maximum and
minimum probability of the words, which appears in the large collection of text. The LDA Technique
is useful in finding the hidden topics for further research in the areas of network and cybersecurity. In
this paper, we collected the abstract of two thousand papers from the Scopus library from 2014 to 2021.
These collected papers are from reputed publications such as Elsevier, Springer, and IEEE Transaction,
etc. The main aim of this study is to find out research trends based on keywords that are untouched or
less research work has been done on that. To the best of our knowledge, this will be the first study done
by using the LDA technique for topic modeling in the areas of network security to demonstrate the re-
search gap and trends for malware and intrusion detection systems.
Chapter 3: A never-ending fight is taking place among malware creators and security experts as the ad-
vances in malware are daunting. The machine learning strategies are indeed the new mode of researching
malware. The purpose of this chapter is to explore machine learning methods for malware recognition
and in general, deep learning methods. The chapter gives complete explanations of the techniques and
resources used in a standard machine learning process for detecting malware. It examines the study is-
sues that are posed by existing study methods and introduces the potential avenues of study in future. By
administering a study to the participants, scholars have a better knowledge of the malware detections.
We start by discussing simple dynamic modelling methods, their importance to the data analytics of
malware, and their implementations. We use open access resources such as virustotal.com which review
sample of dynamic analysis in reality.
xvii
Preface
Chapter 4: Software has been the essential element to computers in today’s digital era. Unfortunately,
it has experienced challenges from various types of malwares which are designed for sabotage, crimi-
nal money-making, and information theft. To protect the gadgets from malware, numerous malware
detection algorithms have been proposed in the olden days there were shallow learning algorithms and
recent years there are deep learning algorithms. With the availability of big data for training of model
and affordable and high-performance computing services, deep learning has demonstrated its superiority
in many smart city applications, in terms of accuracy, error rate, etc. This chapter intends to conduct a
systematic review on the latest development of deep learning algorithms for malware detection. Some
future research directions are suggested for further exploration.
Chapter 5: In the last decades, factories have suffered a significant change in automation, evolving from
isolated towards interconnected systems. However, the adoption of open standards and the opening to the
Internet have caused an increment in the number of attacks. In addition, traditional Intrusion Detection
Systems relying on a signature database, where malware patterns are stored, are failing due to the high
specialization of industrial cyberattacks. For this reason, the research community is moving towards
the Anomaly Detection paradigm. This paradigm is showing great results when it is implemented us-
ing Machine Learning and Deep Learning techniques. This chapter surveys several incidents caused by
cyberattacks targeting industrial scenarios. Next, to understand the current status of anomaly detection
solutions, it analyses the current industrial datasets and anomaly detection systems in the industrial field.
In addition, the chapter shows an example of malware attacking a manufacturing plant, resulting in a
safety threat. Finally, cybersecurity and safety solutions are reviewed.
Chapter 6: Considering the situation when the node which is entering in the network is an a malicious
node then it can check and record the various operations of the network which could be responsible for
some serious security issues. In order to safeguard the network we are hereby proposing an approach
which is in view of the advanced encoding strategy, termed as Convolution coding approach. According
to the requirement of network, the initial bits are assigned and by applying the convolution technique, the
final code is generated for each and every node. Considering the fact that it is a digital method, the codes
can be represented by using the binary number system. All the nodes which are a part of the network will
be having their corresponding final binary code, say C. The verification of each node is carried out by
matching the generated code C with the security code within a particular time period before transmitting
the data. This process enables the detection of malicious or attacker node. Furthermore to enhance the
versatility and execution, the system is organized into clusters.
Chapter 7: In real-scenario, we have a large multicast group where node leaves and joins occur fre-
quently, and also the number of nodes leaving and joining is also not proportionate. Hence Scalable
rekeying process is an important issue that needs to be concerned for the Secured Group communication
for dynamic groups. In basic rekeying scheme which is based on the logical key hierarchy, the rekeying
cost depends on the logarithm of the size of group for a join or depart request by the user. However, the
memory efficiency of this Group Rekeying protocol (GREP) is a huge storage overhead over the system.
We aim to provide a survey of various group key management schemes and then propose an efficient
scalable solution based on linked LKH and the linked List data structure. Results have shown that show
that the Linked LKH algorithm has a very low effective cost for rekeying the LKH as compared to the
basic LKH algorithm i.e. based on the number of new joined and departure requests.
xviii
Preface
Chapter 8: This chapter introduces a new kind of cybersecurity system named Botnet Defense System
(BDS) that defends an IoT system against malicious botnets. This chapter consists of two parts. The
former part describes the concept and design of the BDS. The concept is “fight fire with fire”. To real-
ize the concept, the BDS uses bot technology. The BDS builds a white-hat botnet on the IoT system by
itself and uses it to exterminate the malicious botnets. The white-hat botnet autonomously spreads over
the IoT system and thus drastically increases the defense ability. The latter part explains the strategy of
the BDS. The white-hat botnet is a so-called double-edge sword. It defends the IoT system against mali-
cious botnet but wastes the system’s resources. Therefore, the BDS should strategically use the white-hat
botnet. Some strategies have been proposed. Their characteristics are discussed through the simulation
with the agent-oriented Petri nets.
Chapter 9: The Internet of Things is a cutting-edge technology that is vulnerable to all sorts of ficti-
tious solutions. As a new phase of computing emerges in the digital world, it intends to produce a huge
number of smart gadgets that can host a wide range of applications and operations. IoT gadgets are a
perfect target for cyber assaults because of their wide dispersion, availability/accessibility and top-notch
computing power. Furthermore, as numerous IoT devices gather and investigate private data, they be-
come a gold mine for hostile actors. Hence, the matter of fact is that security, particularly the potential
to diagnose compromised nodes, as well as the collection and preservation of testimony of an attack or
illegal activity, have become top priorities. This article delves into the timeline and the most challenging
security and privacy issues that exist in the present scenario. In addition to this, some open issues and
future research directions are also discussed.
Chapter 10: In recent years, the growth of Brain-Computer Interfaces (BCIs) has been remarkable in
specific application fields, such as the medical sector or the entertainment industry. Most of these fields
use evoked potentials, like P300, to obtain neural data able to handle prostheses or achieve greater im-
mersion experience in videogames. The natural use of BCI involves the management of sensitive users’
information as behaviors, emotions or thoughts. In this context, new security breaches in BCI are offer-
ing cybercriminals the possibility of collecting sensitive data and affecting subjects’ physical integrity,
which are critical issues. For all these reasons, the fact of applying efficient cybersecurity mechanisms
has become a main challenge. To improve this challenge, this chapter proposes a framework able to detect
cyberattacks affecting one of the most typical scenarios of BCI, the generation of P300 through visual
stimuli. A pool of experiments demonstrates the performance of the proposed framework.
Chapter 11: Blockchain since 2009 has been gaining more popularity in various fields to use in numer-
ous applications to overcome the security issues such as privacy, transparency, and mutability of data in
process of data sharing. Process of data sharing has many addressed and unaddressed challenges such
as information encryption and decryption, data authentication, storage security, latency time, transfer
speed of data, detecting malicious nodes, prevent the computer system from attacks, trust in the sharing
process. In this paper, we have reviewed the data sharing paper based on blockchain technology and
presented the analysis of various techniques used in the information sharing process. Our comprehensive
analysis is categorizing in the following areas like incentive mechanism-based work, IoT-based data
sharing, healthcare data sharing, and Internet of vehicle data sharing using blockchain
xix
Preface
Chapter 12: Node capture attack is one of the crucial attacks in wireless sensor network (WSN) that
seizes the node physically and withdraws the confidential data from the node’s memory. The chapter
exploits the adversarial behavior during a node capture to build an attack model. We also propose a Fruit
Fly Optimization Algorithm (FFOA) that is a multi-objective optimization algorithm that consists of a
number of objectives for capturing a node in the network: maximum node contribution, maximum key
contributions are some examples of the same. The aim is to demolish the maximum part of the network
while minimizing the cost and maximizing attacking efficiency. Due to the multi-objective function,
we attain a maximum fraction of compromised traffic, lower attacking rounds, and lower energy cost
as contrasted with other node capture attack algorithms. We have developed an algorithm, which is an
enhanced version of FFOA and has even better efficiency than FFOA
Chapter 13: Brain-Computer Interfaces (BCIs) have experienced a considerable evolution in the last
decade, expanding from clinical scenarios to sectors such as entertainment or video games. Neverthe-
less, this popularization makes them a target for cyberattacks like malware. Current literature lacks
comprehensive works focusing on cybersecurity applied to BCIs and, mainly, publications performing a
rigorous analysis of the risks and weaknesses that these interfaces present. If not studied properly, these
potential vulnerabilities could dramatically impact users’ data, service availability, and, most importantly,
users’ safety. Because of that, this work introduces an evaluation of the risk that each BCI classification
already defined in the literature presents to raise awareness between the readers of this chapter about the
potential threat that BCIs can generate in the next years if comprehensive measures, based on standard
mechanisms, are not adopted. Moreover, it seeks to alert academic and industrial stakeholders about the
impact these risks could have on future BCI hardware and software.
xx
xxi
Acknowledgment
Many people have contributed greatly to this book on Advances in Malware and Data-Driven Network
Security. We, the editors, would like to acknowledge all of them for their valuable help and generous
ideas in improving the quality of this book. With our feelings of gratitude, we would like to introduce
them in turn. The first mention is the authors and reviewers of each chapter of this book. Without their
outstanding expertise, constructive reviews and devoted effort, this comprehensive book would become
something without contents. The second mention is the IGI Global staff for their constant encourage-
ment, continuous assistance and untiring support. Without their technical support, this book would not
be completed. The third mention is the editor’s family for being the source of continuous love, uncon-
ditional support and prayers not only for this work, but throughout our life. Last but far from least, we
express our heartfelt thanks to the Almighty for bestowing over us the courage to face the complexities
of life and complete this work.
Brij B. Gupta
October 1, 2021
1
Chapter 1
Machine Learning for
Malware Analysis:
Methods, Challenges, and
Future Directions
Krishna Yadav
National Institute of Technology, Kurukshetra, India
Aarushi Sethi
National Institute of Technology, Kurukshetra, India
Mavneet Kaur
National Institute of Technology, Kurukshetra, India
Dragan Perakovic
https://orcid.org/0000-0002-0476-9373
University of Zagreb, Croatia
ABSTRACT
Companies and organizations are collecting all sorts of data ranging from nominal feedback like customer
reviews to highly classified data like medical records. With data being such a critical aspect of most
of the operations around us, cybercriminals are looking for an opportunity to misuse this information.
One such device that cybercriminals use to further their malicious intent is malware. Over the years,
these cybercriminals have become immensely powerful using the knowledge of previous attacks. Hence,
malware analysis and methods to troubleshoot the problems arising due to malware attacks is the need
of the hour. Over time, different new approaches have been developed to defend malware. However, in
recent times, machine learning-based malware analysis has gained popularity. The capacity to detect
possible future malware by learning from existing malware patterns makes this method very popular. In
this chapter, the authors have introduced different malware and the machine learning-based approach
that has been developed in recent times to mitigate malware.
DOI: 10.4018/978-1-7998-7789-9.ch001
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Machine Learning for Malware Analysis
INTRODUCTION
Malware or malicious software is an umbrella term that comprises all the software that is designed with
the purpose of corrupting or harming any program, device, service, or network. Once malware pen-
etrates through the network and gets access to the files, it may infect or corrupt the data, steal the data
or even use it for identity thefts. With IoT rapidly becoming a reality, more and more devices are being
connected, which means that if one device in a system gets infected (Sharmeen, Shaila, 2019), all the
other devices are at risk of being infected as well. The trend of the number of major cyberattacks over
the years is pretty sporadic. However, with the introduction of machine learning models, there has been
quite a downfall for the past two years (2019 and 2020), with the drop in the number of malware attacks
going down by 43.3%. These machine learning models range from probabilistic to decision tree-based
to deep neural networks. Over the years, numerous types of malware have been created that work in dif-
ferent ways to harm your files. Knowing what type of malware has infected your network is extremely
helpful in order to find suitable techniques for patching.
There are various kinds of malware available these days. Figure 1 gives a pictorial representation of
different types of malware. One of them is a computer virus (UCCI, Daniele, 2019). Computer viruses
are one of the most common malware. They usually come attached to a file. Once the file is opened, it
corrupts the system by transferring from one programme to another. It can transfer through programmes,
computers, and even networks. Computer viruses can spread by email, text message attachments, audio
files, etc. The next category of malware is worms. Unlike viruses, worms do not require any action by
the victim; they replicate themselves by finding loopholes in the security of the software or operating
system. The main purpose of worms is to destroy the files that it has access to and make them unusable
by using certain encryption or corruption techniques.
2
Machine Learning for Malware Analysis
Further knowledge of specific footprints of each malware can be useful. These footprints or signatures
of a malware lead to the detection of patterns which can be picked up by machine learning algorithms.
This topic is beyond the scope of this chapter but G. Cabau, M. Buhu (2016) can be referred for more
information.
These days every nation is growing rapidly in terms of technologies, transportation, health, etc. Or-
ganizations are competing against each other and are very curious to know what others are doing. Some
organizations develop malware to spy on other systems, which are known as Trojans. Trojans disguise
themselves as harmless files, but once it gains access to your files, it can spy on the victim’s activities,
steal data from the device or crash entire programmes. It can affect the performance of the computer
system by slowing it down, increasing the number of pop-ups, installation of new, unfamiliar software,
increase in the number of spam, etc. Another form of malware is spyware which is not as harmful as
trojan, and the main objective is to spy on its victim’s activities. Even the most menial tasks like the
webpages that a user uses can be highly valuable data for an attacker. Further, it can send back sensitive
financial, medical, or personal information that may be used for fraud, stealing, identity theft, etc. These
software are usually installed onto the devices without explicit permission or knowledge of the user.
When an attack is performed to infect a wide range of systems connected in a network, then botnets
are used. Botnets are malware that taps into the network of connected devices like mobile phones, com-
puters, and other smart devices that are connected through IoT technology. It then gains remote access
to all these devices, and the hacker may use it for malpractices. Several DoS/DDoS attacks in history
have been carried out with botnets. Another type of malware is a rootkit which is programmed in such
a way that it penetrates the network easily to get access to the computer system and then allows access
to other, more dangerous malware. If a computer is infected with rootkit malware, then hackers can
execute commands remotely and change the configuration of the system. As data is a very important
part of any organization, these days, hackers are targeting the data of different organizations. A malware
called ransomware is widely being used to infect a system and encrypt all its data. The data can only be
recovered if the key is entered, which is provided by the hackers if the organization becomes ready to
pay an amount known as ransom to these hackers.
In the rest of the section, section 2 describes various malware detection approaches that have been
developed till now. Section 3 discusses various machine learning algorithms, the feature extraction pro-
cess, and required datasets for malware analysis. Section 4 discusses the challenges involved in using
machine learning algorithms in malware analysis. Section 5 discusses the future directions, and finally,
section 6 discusses possible integration of recent emerging topics in AI for malware detection. Finally,
section 7 concludes the paper.
Since malware can inversely infect the health of a computer system or network, it is important to invest
in quality anti-virus softwares. These anti-virus softwares are based on the two major malware detec-
tion methods namely - signature based detection and heuristic based detection. Another technique for
memory based detection is also being developed, but it is a relatively novel concept.
3
Machine Learning for Malware Analysis
Currently, the majority of the antivirus available use signature-based methods for malware detection
(Sathyanarayan, V. Sai, 2008). This technique comprises capturing a typical signature from the malware
and using this knowledge to handle similar malwares. A signature is a unique footprint or pattern used
by a malware to attack the system. This may be a unique sequence of bytes or a file hash. This technique
has been used since the development of the first antivirus softwares, but it has its share of advantages
and disadvantages. One of the major advantages that play a pivotal role in making this technique so
popular is that it is very fast and efficient. The pre-existing knowledge of this footprint also factors into
a low false positive rate in general. The main disadvantage of such an approach is that it does not work
well with newly released malware, and it is not effective against obfuscation. Obfuscation (Wroblewski,
Gregory, 2002) is a process adopted by attackers to render the pre-existing code illegible or hard to read
while maintaining its functionality. It prevents the malware from being detected but at the same time
disrupts the execution of the victim’s programme. Some of the obfuscation techniques are - changing the
look of the programme while keeping the same functionality using dead code insertion, integrating the
malware in the victim’s programme using code insertion methods, switching or reassigning the registers.
Heuristic based detection techniques were developed as a response to signature based detection tech-
niques, specifically its shortcomings (Bazrafshan, Zahra, 2013). Unlike signature based detection, which
is developed on the concept of comparison, heuristic based detection mechanisms run the files in a care-
fully orchestrated and/or monitored environment to detect anomalies and malwares. It is also effective
against polymorphic viruses because of the aforementioned property. In this method, the behaviour of
the malware is analyzed during the training or learning process, and the files are categorized as mali-
cious or legitimate based on the learning. This is the testing phase. This concept of training and testing
incorporates well with the training-testing structure of machine learning. The ability to learn and adapt
gives this technique an edge when it comes to zero-day malwares. However, the same property can be
disadvantageous as well as it has higher false positive rates. This technique also takes way longer than
signature based detection.
Heuristic based detection makes use of machine learning models to predict the malicious or corrupt
components. This process and steps involved in it are discussed in great detail.
Since malware is a diverse term, machine learning can be used to uncover various hidden patterns, and
various deductions can be made by making use of the same data but different features. Machine learning-
based malware analysis comprises three different steps, which are represented in figure 2. In the first
step, i.e., Malware detection (Idiaka, Nwokedi, 2007) and (Siddiqui, Muazzam, 2008), the goal is to
determine whether a given sample is malicious or benign. Such models form the basis of many complex
machine learning solutions as the very first step to analyse malware analysis is determining whether a
file poses any threat to the security of a system. Malware category detection (Siddiqui, Muazzam, 2008)
is determining what type of malware has been encountered. This is one of the most important types of
4
Machine Learning for Malware Analysis
malware analysis as each malware needs to be handled differently depending on its mode of operation and
method of attack. Each malware has different footprints which can be utilized to detect its type. Malware
comparison, as the name suggests, is comparing how different or similar a sample is to a certain type
of malware or is it a variant of a type of malware or not. It is especially useful against new malwares as
any similarity to certain malware may be useful for developing a patch against it.
In machine learning, the dataset is used to train a model using an algorithm that helps in predicting
the outcome of a proposed sample. This is a little different than rule-based classification because the
model is not a set of predefined rules, rather it is adaptive, and it learns with each new sample. Given
the objectives, classification models (Liu, 2017) are a clear choice. The classification can either be
malicious or benign files for malware detection, classifying the malwares into types or simply classify-
ing a sample into high, medium, low categories based on its similarity or difference to specific type of
malware. The possibilities are endless, and the categories can be modified based on the need. However,
the three examples mentioned above are by far the most common types of classification prevalent in
researches that focus on malware analytics. The three main types of machine learning are - supervised,
unsupervised and reinforcement learning, out of which the first two are more popular at the moment.
Supervised machine learning comprises a set of training and testing data with the output of the training
data labeled. Whereas in unsupervised machine learning, the output sets are not labeled. Many machine
learning algorithms and their effectiveness in this scenario are discussed below. These techniques include
KNN, SVM, Logistic regression, Random Forest, Naive-Bayes, Gradient Boosting decision trees, etc.
for supervised machine learning and for unsupervised machine learning, k-means clustering, hierarchical
clustering, etc. are described. Apart from these machine learning algorithms, deep learning algorithms
like Artificial Neural Network, Convolutional Neural Network, and Recurrent Neural Network are being
used for complex purposes like malware similarity and difference as they are efficient when it comes to
working with non-linear, images or sequential data.
A few common machine learning algorithms and how they are used for malware detection are dis-
cussed down below.
5
Machine Learning for Malware Analysis
1. KNN
This is a supervised machine learning algorithm. KNN or K Nearest Neighbours model is based
on the concept of clustering using distance. In this method, imagine the features to be the axes of a
multidimensional plane. Each of the samples is plotted on the multidimensional plane. The prox-
imity of these samples is used to define their clusters or categories in which the output is divided.
The criteria of proximity is calculated by the distance. The distance of a sample to the cluster is
measured between the sample and the centroid of the cluster, which is also called the cluster head.
In this method, K i.e., the number of neighbours to be considered for assigning a cluster to a given
sample, distance, i.e., the type of distance- euclidean, manhattan etc. are some of the hyperparam-
eters that need to be tuned to get the best results. Since this process has the ability to divide the
samples into multiple clusters, it can be used to determine the type of malware (Choi, Sunoh, 2020).
2. SVM
SVM comes under the category of supervised machine learning algorithms. SVM or Support
Vector Machine is based on the concept of defining a hyperplane and classifying the samples into
categories based on which side the sample falls on with respect to said hyperplane. Similar to
KNN, the samples are plotted on a multidimensional plane, and optimized hyperplanes are plotted
to classify the samples into categories. The division is done into multiple categories with the help
of a kernel. A kernel is a method for using a linear classifier to solve non-linear classifications.
In SVM, the main hyperparameters are type of kernel- polynomial, linear, rbf, etc and C, i.e., the
penalty parameter. This is also a method capable of effective multi-class classification and can be
used for malware type, similarity, difference, etc.
3. Logistic Regression
This is a supervised machine learning algorithm. Logistic Regression is a model that is more often
than not used for binary classification. It predicts the output in the form of the probability of class
0 using the log probability formula. Since this model is generally used for binary classification, it
can be effective only in malware detection of benign and malicious classes.
4. K-Means Clustering
K-means clustering is loosely related to K Nearest Neighbour, i.e., it classifies a sample based on
its proximity with the centroid of a cluster, but this is an unsupervised machine learning algorithm,
i.e., inferences are only made based on the input vectors and not the labels assigned to them. In
this algorithm, K, which is the number of clusters, is a hyperparameter. Since output labels are
not necessary for such algorithms to function, such models can be effective in scenarios where
the output is uncertain or unavailable. Such scenarios are common in context to new variants,
and types of malwares as their effect on the environment have not already been analyzed properly
(Feizollah, Ali, 2014).
Other popular algorithms are naive-bayes, decision trees based, ensemble, hierarchical clustering, etc.
All of these algorithms can be useful in various aspects of malware detection using machine learning.
Algorithms like KNN and SVM are extremely effective when it comes to the types and variance in mal-
ware. Logistic Regression comes in handy when we want to differentiate between malicious and benign
files. Unsupervised learning algorithms like K-Means clustering can be used for critical and unknown
files. Authors at (Shhadat, Ihab, Amena Hayajneh, 2020) have used different machine learning algorithms
for malware detection on the same benchmark dataset. The results obtained is presented in table 1.
6
Machine Learning for Malware Analysis
Although machine learning algorithms are very successful in detecting malware, these days, dif-
ferent deep learning algorithms like CNN(Convolution Neural Network) and autoencoders have also
been very efficient. Authors at (Ganesh, Meenu, 2017) have presented that information can be obtained
from the application of smartphones to and can be converted to images for applying CNN. They have
scanned several apk files and decompiled using a parser and disassembler. The disassemble file is a
.xml file which is later used to make an image of vector size 12*12. This image acts as an input for the
CNN algorithm. They have obtained an accuracy of 95%. Finding good features to prepare a dataset
to train deep learning algorithms can sometimes be time-consuming. Additionally, ineffective features
might be generated through some poor feature engineering processes. To solve the problem, authors at
(M. Yousefi-Azar, 2017) have used autoencoders to generate machine-generated features. Autoencoder
generates features by distributing the dataset into some lower-dimensional space and trying to reconstruct
the original data from the learned features set. Authors, with the help of experiments, have demonstrated
that unsupervised algorithms like autoencoders can generate better discriminative features than traditional
feature engineering processes.
After defining a model and compiling it, it is trained on the dataset. After training the model, it can
be used to predict the outcome.
Figure 3 represents an end-to-end approach (Kaspersky, 2021) to using machine learning to build a
model for malware analysis. Usually, a large amount of raw data needs to be extracted and processed if
the process of analyzing starts from scratch. The raw data is still in the form of complex data structures.
These features can be sandbox logs, disassembler logs, etc. These features can be extracted in a pre-
execution phase and should be done in a high-computation environment. These ‘heavy-computation’
features are highly effective in predicting the presence or absence of a model in file. A large-scale model
can be created that predicts the same. The large scale or general model is used to divide the files into
harmful and benign. Once they are broadly categorised, ‘lighter’ features from the suspicious files are
extracted for further inspection. The light-computation features are fille structure, content statistics, API,
etc. The outcome of these models can be used for specific, more objective driven purposes like detecting
the type or class of malware, whether it is a variant or not, etc., to make a new model. Once this model
is created and trained, it can be used for the purpose of prediction. This model is updated regularly and
is adaptable as new malware requires new information. Machine learning provides a holistic product
as our model is not just dependent on predefined information. It learns as new files are encountered.
7
Machine Learning for Malware Analysis
Datasets
The datasets used for malware can be structured in such a way that they classify a sample as malicious or
benign, or it can be used to classify the type of malware that has attacked/ is being observed. The datasets
can be obtained from public sources, security setup providers, sandboxed analysis service providers like
Malware DB, VX Heaven, or VirusShare repository. Various organizations do machine-learning-based
malware analysis by creating their own datasets. Due to such problems, there are very few malware based-
datasets available publicly. However, institutes like UNB have made various efforts to publicly share
such datasets. Malware in android mobiles is increasing at an alarming rate. Considering this situation,
UNB has developed CCCS-CIC-AndMal-2020 dataset that comprises 200K benign and 200K malware
samples that further includes 14 prominent malware with 191 eminent malware families (D. Sean, B. Li,
2021). The CICMalDroid 2020 is a dataset that contains over 17,341 malware samples (S. Mahdavifar,
A.Firtriah, 2020). The dataset was collected between December 2017 to December 2018. It contains
five different categories of malware, i.e., Adware, Banking malware, SMS malware, Riskware and Be-
nign. CIC-InvesAndMal2019 is another malware dataset created that includes permissions and intents
as static features and API calls along with generated log files as dynamic features (Laya Taheri, 2019).
Besides those features, they have also added features such as battery states, packages, process logs, log
states, etc. Another CIC-AndMal2017 is an android-based malware dataset that contains more than 10,
854 samples (A.H Lashkari, 2018). Researchers have collected around six thousand apps from google
play and extracted the features that contain four different malware classes, i.e., Adware, Ransomware,
Scareware, SMS Malware.
If large companies with huge data repositories are developing a model, they prefer to use the in-house
data to train the model while simultaneously keeping a check on the security of their own network. E.g.,
Microsoft created a dataset of more than 500,000 malicious files and shared it as a public dataset. The
most popular types of datasets are malware detection datasets that classify a given sample as malicious
or benign (A.H Lashkari, 2018). The benign samples are the harmless, legitimate files or programmes.
However, Internet Service Providers (ISPs), honeypots, and Computer Emergency Response Teams
(CERTs) developed both benign and malicious datasets for the public in order to analyze the other aspects
and properties of malware as well. When it comes to malware detection datasets, the ratio of number
of benign samples to number of malware samples is quite large, which may lead to the development of
8
Machine Learning for Malware Analysis
a biased model. To overcome this problem, techniques such as undersampling or oversampling (Pang,
Ying, 2019), the latter more than the former, may be made use of to balance the number of samples.
Authors at (D.Gavrilut, 2009), have categorized malware dataset into, training, test and scale-up dataset.
The categorized dataset is represented in table 2.
Feature Extraction
When it comes to extracting features (Sihawal, 2018) that are best suited for a specific problem, there
are three types of feature extraction methods - static, dynamic, and hybrid analysis.
Static
For such analysis, PE or Portable Executable files do not need to be run in order for features to be ex-
tracted from them. For a PE file to be examined and analysed for feature extraction, tools like IDA Pro,
OllyDbg, Exeinfo PE, PEstudio are used to disassemble them and extract information about the malware,
receive instructions and pick up on patterns. Some of the static features that are commonly used are API
calls, string signature, control flow graph (CFG), opcode (operation codes) frequency, and byte sequence
n-grams. These have been discussed further in detail later on.
Dynamic
In such analysis, the PE files do not need to be disassembled for analysis. They are analyzed by running
them in a controlled, monitored environment. This is really effective against obfuscated and polymorphic
or zero-day viruses that use methods that are different or variations of pre-existing malwares. However,
as the model is constantly being developed or updated, it takes a long time and a lot of computation to
train the model. Hence, such a setup can be problematic in time sensitive environments. Features like
API calls, API sequences or file system registry can be obtained using dynamic techniques like func-
tion call monitoring, function parameter analysis, instruction traces, and information flow tracking. As
mentioned before, for dynamic analysis the files are monitored in a controlled and monitored environ-
ment. This environment can be generated with the help of emulators, debuggers, simulators or virtual
machines. The challenges faced during the creation of the environment are discussed in a later section.
9
Machine Learning for Malware Analysis
Hybrid
Hybrid analysis is performed by making use of both, static and dynamic features. This approach has been
proved more effective than static and dynamic analysis as it makes use of the advantages and disadvan-
tages of static and dynamic features to create a balanced model for the proposed problem.
Features Description
A good set of features can boost the accuracy of machine learning algorithms for malware analysis. The
popular features (11) used for malware analysis are discussed below.
Byte Sequence
Byte sequence is a static feature. While searching for footprints of a malware, specific byte sequences
that are found in a sequence can lead to the recognition of a corrupted or malicious file. The distinction
can be based on the byte size or certain characteristic strings called n-grams.
N-Gram
N-gram is a static feature. It is a subsequence of a prominent sequence that has been divided into chunks
of size n. The number of n-grams that are present in a sequence are extracted as a feature for analysis.
Malwares can be identified based on the number of n-grams found in a suspicious sample in combination
with other features. N-grams of size no more than three are preferred by researchers.
Opcode
Opcodes are extracted through static analysis. Opcodes or operation codes are designed to identify the
machine level operations that are executed by the PE files. Opcodes are a very common feature as they
extract the frequency of opcodes in a sample or its similarity and difference to another sequence to detect
the presence or absence of malware. These machine level operations can further be divided into math-
ematical instructions, program sequencing and control instructions, instructions for accessing memory etc.
String
This feature can be extracted through static analysis. While looking for the presence of a malware,
characteristic strings like author name, file names, typical code fragments, etc., are searched in order
to identify the footprint.
Network Activity
This feature comes under the category of dynamic features. Network analysis can reveal a lot of important
information. Monitoring used protocols, HTTP/HTTPs requests, security firewalls, TCP/UDP ports, etc.,
can be useful in determining the conditions where the health of a computer network is compromised.
10
Machine Learning for Malware Analysis
The incoming and outgoing traffic of a sample can be monitored in order to be alerted if suspicious
fluctuations are taking place.
This is one of the few features that can be extracted both through static and dynamic analysis. These
can be extracted through disassembling the PE file and obtaining the information about all the calls
that are going to be executed once the file is run, or it can be run in a secure, observed environment to
get the knowledge of the execution calls. The amount of data that is obtained by extracting this feature
is massive, and therefore often, preprocessing techniques are used to break it down to simpler formats
that are easy to access and read. One of the more common formats is control flow graph. A control flow
graph (CFG) is a directional graph that depicts the flow of calls in a programme. Like a usual graph it
uses nodes and edges to convert the list of calls into a graphical format.
CPU Registers
This feature comes under the category of static feature analysis. It can be extracted such that the values
stored in a register, the number of registers, the re-assignment of registers, or the presence of a hidden
or unwanted register can be recorded.
This can be a static or dynamic feature. It is important to have knowledge about which files are run-
ning which operational commands and the interaction of these commands of a particular sample with
the environment. The information that can be mined from this data covers a lot of ground. It includes
information like how a file is read or modified in the environment, which files belong to which directory
or repository, the frequency of presence of a sample file in infected and non-infected systems, how a
certain sample is deleted from the environment. This can easily be extracted while a simulated or virtual
environment is created using sandbox services and memory analysis toolkits. Using static analysis, the
PE files can be examined for sections, imported symbols, or compilers preferred for execution.
In recent developments, another subset of Machine learning - reinforcement learning is becoming
popular for feature selection. This is an upgrade from the traditional machine learning techniques that
require human input to function. Furthermore, these techniques only help in detecting single view fea-
tures. With Reinforcement learning, multiview features can be taken into consideration and variants of
multiple malware can be detected. (Z. Fang, 2019) proposes a Deep Q-learning based Feature Selection
Architecture that trains its model on the basis of its interaction with the feature space.
These are just a few features that can be used for preparing a dataset for the model to train on. Several
specific features can also be generated using machine learning techniques that are in relevance to certain
objectives. Keep in mind that too many features can overcomplicate things and the model might not
fit onto generic data. On the other hand, less features can lead to lack of information for the model to
extract patterns. Hence, it is vital for us to select the appropriate number of features to be used. Feature
selection techniques like correlation come handy in this scenario.
11
Machine Learning for Malware Analysis
Model Evaluation
In highly sensitive matters like that of malware detection, it is very important to have a good, appropri-
ate model as even one corrupt file can infect a lot of valuable data. Here are few of the criteria a model
should fulfil in order for it to be called good.
It is very important for our model to be trained on a large dataset to have high accuracy. If our dataset is
not large enough, it might now be able to extract enough information to build a model that does not fit
well with the training or testing dataset. Hence, our dataset should be large enough to contain diverse
cases that help in not developing a bias towards specific recurring cases. However, our dataset should
also not have too many features that may lead to overfitting. In such a case, our model does well with
the training dataset but doesn’t do good enough with the testing dataset. Hence our dataset on which we
train our model should be optimum.
Since new malwares are being developed by writers everyday, it is important for our model to respond
properly to the latest attacks. Not only are there new types of malware, but also companies are producing
new benign files to be up-to-date with the latest technology. Hence, it is also important for our model to
recognize a new benign sample as much as it is to recognize a malicious one. The algorithms should be
adjustable and flexible and our model needs to be adaptable to new changes while retraining. The new
datasets must also contain the latest samples so that the model can extract information and learn from
them while retraining.
When a harmless file is labeled as malicious, that outcome is labeled as a false positive. False positives
may occur under various circumstances like our model is not trained on a proper dataset, or new types
of benign files are being generated and fed to the model that it is unable to recognise and label properly.
A false positive may end up being as harmful as a malicious file as a lot of important information may
be lost due to one mis-classification. Therefore, It is important for a model to have a low false rate and
in an ideal scenario it tends to zero.
Machine learning-based malware analysis has brought a great revolution in detecting malware; however,
it is still not fully developed. There are many challenges associated with this approach. Some of them
are described below.
12
Machine Learning for Malware Analysis
Since malware authors are aware of malware detection softwares, they are actively using the knowl-
edge to create techniques that help a malware go unnoticed by them (Gibert, 2020). For static analysis,
they prefer techniques like obfuscation, packaging or encryption to hide the information regarding the
malware. A solution to these techniques is dynamic analysis. During dynamic analysis these files can
be de-obfuscated, unpacked or decrypted while being monitored in a controlled environment. A more
advanced anti-analyzation technique is environmental awareness. In this technique, the malware is de-
signed such that it knows when the file is being run for inspection. This is most commonly prevalent in an
environment that is set up on virtual machines, sandboxes etc. due to the files that are installed for their
setup or the mode that they run in. The malware instantly recognizes when it is being run in a monitored
environment and stops the execution of malicious payload. An alternative approach can be time-based
execution. In this technique the malicious payload is set up such that it only runs on specific dates and
time. Another option for malware authors is user interaction based. In this technique, the malicious
payload is only run when a user interacts with a file in a previously defined manner. These challenging
techniques need to be addressed head on while malware analysis as they are becoming more and more
popular among attackers due to their success rate. One counter-technique identified by researchers is
by using binary reverse engineering such as function similarity identification to realize the presence of
de-obfuscating, unpacking or decrypting techniques.
Operations such as Opcodes, instructions, APIs and system calls, etc., are valuable features for the
prediction of a malicious sample. But by disintegrating raw data into simpler data structures, the out-
come may result in a large number of features to be trained on. If the number of features go beyond an
optimum number, our model will not turn out to be generalized enough and may suffer from what is
called the ‘curse of dimensionality’. At the same time, losing some of the features may lead to loss of
important information and resulting in error. This challenge can be addressed by using better feature
selection techniques or changing the type of algorithm used, but sometimes it is not feasible to change
the initial structure of a model. In such cases program analysis advances can be used to improve the
accuracy of the disassemblers and decompilers as they have a higher error rate during dynamic analysis
than the operations during the static analysis. Hence methods that improve the quality of disassemblers
and decompilers can be used. A hybrid of both can also be considered for improving the overall quality.
FUTURE DIRECTION
Nowadays malware authors are using auto-generative tools to build out new malware and their vari-
ants at a rapid speed (Zarras, 2014). This is quite worrisome for the organizations dealing with critical,
sensitive and high frequency data as the frequency of input might not give enough time for a solution to
a day zero malware. A proposed solution to this problem can be building models advanced enough to
suggest solutions to a given problem. This is a predictive maintenance solution as it allows not only the
13
Machine Learning for Malware Analysis
file to be separated but also a possible method to navigate through a corrupted system. Further, public
sandboxes and repositories can be used to keep the dataset updated with the latest malware that has been
detected. This solution is the one that can be implemented in the near future as the field of predictive
maintenance still needs a lot of development before it can be considered as a reality.
Malware Attribution
Another method that is gaining momentum these days is the method of identifying the malware by the
author of the file, i.e., the attribution of a technique to its author (Gupta, 2019). It is quite an unusual
method, but researches have shown that the possibilities of the type of malware can be streamlined to a
great extent if the author of a malicious code is revealed. The used programming language, IP addresses,
and URLs that have been included or the style of the comments and/or code that has been used to develop
a malicious payload can be used to identify the author (Gupta, Brij B, 2021). Some authors want to be
boastful of their developments and try to leave an encrypted signature behind in their codes. These can
be some useful clues to figure out the author or the ATP group that is behind certain malware. However,
the dependence of this technique on finding the source code using leaks or public disclosures makes
it unreliable, at least in the near future, but once a concrete development is made in this direction, this
method can be revolutionary in the field of malware analysis.
Generating variants is a simpler task for attackers in comparison to developing a whole new malware
as it requires just a few changes to develop a new, stronger variant of a malware (Yamaguchi (2021)).
Attackers use this to their advantage as it enables them to continuously attack their target in rapid suc-
cession. This may lead to weakening of the strength of the security system of an organization and at the
same time corruption of important files. Researchers are developing solutions using machine learning
that aim at being effective against all the possible variants of a malware without necessarily being de-
veloped first (Gou, Z, 2017).
While all malwares have malicious intentions, some are more harmful to others. Hence, it is important to
determine the level of urgency for each malware especially when dealing with sensitive data. The level
of urgency can be based on several factors like extent of damage, number of days since it was launched
or other local factors specific to the data. This can lead to faster results and less computational costs.
While this is a far-fetched trend, researchers are trying to figure out a solution that helps in preventing
any sort of malware from attacking. Discovering a common factor would mean that a common solution
for all the malware would be possible (Gou, Z, 2017). Such solutions also have a better chance at be-
ing adaptive to new malware. However, this is still far from reality as the types of malware discovered,
all of them have quite distinctive footprints and ways of corrupting computer networks. None-the-less,
researchers are hopeful regarding this approach.
14
Machine Learning for Malware Analysis
A rapidly growing niche of machine learning is deep learning. Deep Learning networks make use of
artificial neural networks that are specialized in picking up on complex, non -linear patterns that machine
learning algorithms may miss out on. CNN can be used to project the features as an image and predict
the outcomes. RNN may be used to suggest the next step in a sequence of attacks by the malware. Fur-
thermore, advanced ML techniques like reinforcement learning are becoming popular as, if successful,
they are better at handling sensitive files.
Internet of Things and Cloud Computing have led to a revolutionaliation of data and how it is collected.
However, since most of the passage of data in IoT is via the internet, the chances of being attacked
by Malware have become higher. In techniques like federated learning, where the updated weights, if
corrupted, can ruin the final result of other devices as well. Hence, malware analysis systems that use
machine learning should be established along with IoT devices for their protection. Furthermore, the
concept of IoT can itself be used to advance the results of malware analysis. Federated learning can be
used to collect the information about other attacks as well that will help in building a better, diverse
model. (Zekri, 2017) describes in great detail the DDoS or Distributed Denial of Service Attacks that
use multiple attackers to attack innocent computers, capture them using bugs and security discrepancies
and then use the majority of the victim’s cloud bandwidth to send a large amount of packets and data.
Though this signature can be detected quite easily using prior knowledge if the malware is present in
native form, zero-day malware goes undetected if machine learning is not used instead of conventional
detection methods.
CONCLUSION
Since data is becoming such an important commodity, attacks on data are also becoming more severe and
destructive. We have learnt about malware, malware detection methods and types of malware. Further
we have discussed some feature extraction techniques followed by a brief introduction of some of the
important features. Some of the popular machine learning algorithms have been discussed in the further
section. It is followed by some pointers that describe the qualities of a good machine learning model that
have to be kept in mind while constructing one. It is important to be aware of some of the difficulties
one might face while attempting to build a machine learning model for malware analysis, so in the next
part the challenges and issues have been described. Lastly, up and coming trends have been featured in
the last section to make the readers aware of the latest developments in the field of machine learning
and malware analysis. Malware analysis is a relatively new concept and a lot of discoveries are yet to
be made in this field and with the help of machine learning we can take this concept one step further.
15
Machine Learning for Malware Analysis
REFERENCES
Cabău, G., Buhu, M., & Oprişa, C. (2016). Malware classification using filesystem footprints. 2016
IEEE International Conference on Automation, Quality and Testing, Robotics (AQTR), 1-6. 10.1109/
AQTR.2016.7501294
Choi, S. (2020). Combined kNN Classification and hierarchical similarity hash for fast malware detec-
tion. Applied Sciences, 10(15), 5173.
Fang, Z., Wang, J., Geng, J., & Kan, X. (2019). Feature Selection for Malware Detection Based on
Reinforcement Learning. IEEE Access: Practical Innovations, Open Solutions, 7, 176177–176187.
doi:10.1109/ACCESS.2019.2957429
Feizollah, A. (2014). Comparative study of k-means and mini batch k-means clustering algorithms in
android malware detection using network traffic analysis. In 2014 international symposium on biometrics
and security technologies (ISBAST). IEEE.
Ganesh, M. (2017). CNN-based android malware detection. In 2017 International Conference on Soft-
ware Security and Assurance (ICSSA). IEEE. 10.1109/ICSSA.2017.18
Gavriluţ, D., Cimpoeşu, M., Anton, D., & Ciortuz, L. (2009). Malware detection using machine learn-
ing. 2009 International Multiconference on Computer Science and Information Technology, 735-741.
10.1109/IMCSIT.2009.5352759
Gibert, D., Mateu, C., & Planes, J. (2020). The rise of machine learning for detection and classification
of malware: Research developments, trends and challenges. Journal of Network and Computer Applica-
tions, 153, 102526. doi:10.1016/j.jnca.2019.102526
Gou, Z., Yamaguchi, S., & Gupta, B. B. (2017). Analysis of various security issues and challenges in
cloud computing environment: a survey. In Identity Theft: Breakthroughs in Research and Practice (pp.
221-247). IGI global. doi:10.4018/978-1-5225-0808-3.ch011
Gupta, B. B., & Sheng, Q. Z. (Eds.). (2019). Machine learning for computer and cyber security: prin-
ciple, algorithms, and practices. CRC Press. doi:10.1201/9780429504044
Gupta, B. B., Yadav, K., Razzak, I., Psannis, K., Castiglione, A., & Chang, X. (2021). A novel approach
for phishing URLs detection using lexical based machine learning in a real-time environment. Computer
Communications, 175, 47–57. doi:10.1016/j.comcom.2021.04.023
Idika, N., & Mathur, A. P. (2007). A survey of malware detection techniques. Purdue University.
Kaspersky. (2021). Machine Learning for Malware Detection. Retrieved from https://media.kaspersky.
com/en/enterprise-security/Kaspersky-Lab-Whitepaper-Machine-Learning.pdf
Keyes, L., & Kaur, L. Gagnon, & Massicotte. (2021). EntropLyzer: Android Malware Classification and
Characterization Using Entropy Analysis of Dynamic Characteristics. In Reconciling Data Analytics,
Automation, Privacy, and Security: A Big Data Challenge (RDAAPS). IEEE.
Kim. (2018). A multimodal deep learning method for android malware detection using various features.
IEEE Transactions on Information Forensics and Security, 14(3), 773-788.
16
Machine Learning for Malware Analysis
Lashkari, A. H. (2018). Toward developing a systematic approach to generate benchmark android mal-
ware datasets and classification. In 2018 International Carnahan Conference on Security Technology
(ICCST). IEEE. 10.1109/CCST.2018.8585560
Liu, L., Wang, B., Yu, B., & Zhong, Q. (2017). Automatic malware classification and new malware
detection using machine learning. Frontiers of Information Technology & Electronic Engineering, 18(9),
1336–1347. doi:10.1631/FITEE.1601325
Mahdavifar, Fitriah Kadir, Fatemi, Alhadidi, & Ghorbani. (2020). Dynamic Android Malware Category
Classification using Semi-Supervised Deep Learning. The 18th IEEE International Conference on De-
pendable, Autonomic, and Secure Computing (DASC).
Pang, Y. (2019). A signature-based assistant random oversampling method for malware detection. In 2019
18th IEEE International conference on trust, security and privacy in computing and communications/13th
IEEE international conference on big data science and engineering (TrustCom/BigDataSE). IEEE.
10.1109/TrustCom/BigDataSE.2019.00042
Sathyanarayan, V. S., Kohli, P., & Bruhadeshwar, B. (2008). Signature generation and detection of mal-
ware families. In Australasian Conference on Information Security and Privacy. Springer. 10.1007/978-
3-540-70500-0_25
Sharmeen, S., Huda, S., Abawajy, J. H., Ismail, W. N., & Hassan, M. M. (2018). Malware threats and
detection for industrial mobile-IoT networks. IEEE Access: Practical Innovations, Open Solutions, 6,
15941–15957. doi:10.1109/ACCESS.2018.2815660
Shhadat, I., Hayajneh, A., & Al-Sharif, Z. A. (2020). The use of machine learning techniques to ad-
vance the detection and classification of unknown malware. Procedia Computer Science, 170, 917–922.
doi:10.1016/j.procs.2020.03.110
Siddiqui, M., Wang, M. C., & Lee, J. (2008). A survey of data mining techniques for malware detection using
file features. Proceedings of the 46th annual southeast regional conference. 10.1145/1593105.1593239
Sihwail, Omar, & Ariffin. (2018). A survey on malware analysis techniques: Static, dynamic, hybrid and
memory analysis. International Journal on Advanced Science, Engineering and Information Technol-
ogy, 8(4-2), 1662.
Taheri, A., & Lashkari. (2019). Extensible Android Malware Detection and Family Classification Us-
ing Network-Flows and API-Calls. The IEEE (53rd) International Carnahan Conference on Security
Technology.
Ucci, D., Aniello, L., & Baldoni, R. (2019). Survey of machine learning techniques for malware analysis.
Computers & Security, 81, 123–147. doi:10.1016/j.cose.2018.11.001
Wroblewski, G. (2013). General method of program code obfuscation. In The 5th Conference on Infor-
mation and Knowledge Technology. IEEE.
Yamaguchi, S., & Gupta, B. (2021). Malware threat in Internet of Things and its mitigation analysis. In
Research Anthology on Combating Denial-of-Service Attacks (pp. 371–387). IGI Global. doi:10.4018/978-
1-7998-5348-0.ch020
17
Machine Learning for Malware Analysis
Yousefi-Azar, M., Varadharajan, V., Hamey, L., & Tupakula, U. (2017). Autoencoder-based feature learn-
ing for cyber security applications. 2017 International Joint Conference on Neural Networks (IJCNN),
3854-3861. 10.1109/IJCNN.2017.7966342
Zarras, A. (2014). Automated generation of models for fast and precise detection of HTTP-based mal-
ware. In 2014 Twelfth Annual International Conference on Privacy, Security and Trust. IEEE. 10.1109/
PST.2014.6890946
Zekri, El Kafhali, Aboutabit, & Saadi. (2017). DDoS attack detection using machine learning techniques
in cloud computing environments. . doi:10.1109/CloudTech.2017.8284731
18
19
Chapter 2
Research Trends for Malware
and Intrusion Detection
on Network Systems:
A Topic Modelling Approach
Govind P. Gupta
https://orcid.org/0000-0002-0456-1572
National Institute of Technology, Raipur, India
Sanjay Kumar
National Institute of Technology, Raipur, India
ABSTRACT
With more uses of internet-based services, the risk of cyberattacks is growing continuously. To analyze
these research trends for malware and intrusion detection, the authors applied the topic modeling approach
in the study by using the LDA (latent dirichlet allocation) and calculating the maximum and minimum
probability of the words, which appears in the large collection of text. The LDA technique is useful in
finding the hidden topics for further research in the areas of network and cybersecurity. In this chapter,
they collected the abstract of two thousand papers from the Scopus library from 2014 to 2021. These
collected papers are from reputed publications such as Elsevier, Springer, and IEEE Transactions. The
main aim of this study is to find research trends based on keywords that are untouched or on which less
research work has been done. To the best of the authors’ knowledge, this will be the first study done by
using the LDA technique for topic modeling in the areas of network security to demonstrate the research
gap and trends for malware and intrusion detection systems.
DOI: 10.4018/978-1-7998-7789-9.ch002
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Research Trends for Malware and Intrusion Detection on Network Systems
INTRODUCTION
According to a report of Equinix (Catalin Cimpanu et al. 2020), a 45% rise in cyber-attack may be seen till
2023 on the network communication system domain with the increasing use of computer communication
networks and internet-based services. The threat of attack is ever-expanding over network infrastructure
as the different techniques of attacks are being used by attackers. The traditional method of detecting
new attacks is not effective as of as it should be to detect and respond before misuse of unauthorized
users to the computer resources and private information of the organizations. There are various kind
of attack exist known as intrusion, anomaly, malware, viruses, ransomware, adware, Trojan horses,
DDoS, DoS, and many more. Malware is the malicious code that moves across the computer network
to gain unauthorized access into the computer’s critical resources, files, root directory, etc. Malware is
designed to steal information, provide losses for the computer system, embedded into the user’s code
to scan sensitive information, and sending data on third parties servers. Different types of Malware has
been classified based on their activity into computer systems such as adware, Trojan, bot, worm, virus,
spyware, Ransomware, Rootkit, downloader, Launcher, Backdoor, etc. (Gibert et al., 2020), malware
analysis follows two approaches for detecting malware as static analysis and dynamic analysis. The
static approach follows the finding unwanted pattern without executing the codes, whereas the dynamic
approach works on running code and monitoring the behavior of systems activity (Ren et al., 2019).
Intrusion is unauthorized access that tries to intrude into the privacy of a network. There are different
types of intruders such as masquerade and clandestine users to detect intruders, an in-network system
deploys a smart system known as an intrusion detection system. An intrusion detection system (IDS) is
a security model of the network system used to trace the unauthorized activity of the network through
the scanning process of traffic analysis of network packets. The intrusion model identifies the system
activities behavior whether it is normal or abnormal and responds to it to the network administrator. The
IDS works based on predefined records it contains and finds out the intrusion. It is a predictive model
used in the cyber and network security domain that consists of various machine learning algorithms in
identifying the intrusion very accurately (Belavagi et al., 2016).
There is two class of intrusion detection systems such as network based detection system and host
based detection system (Farnaaz et al., 2016). Network-based detection systems work as passive mode
means only monitor the activity of network traffic and send it to central authority while in the case of
host-based systems, it is capable of detecting the internal activity of the system and also filter the packet
data. The IDS can be classified as anomaly-based and misuse-based detection to detect new attacks and
unknown patterns as well as known attacks from databases respectively (Stein et al., 2005, Belavagi
et al., 2016). There is a lot of work is being done by researchers in the field of intrusion and anomaly
detection and proposed various techniques of machine learning, deep learning, and data mining tools in
the areas of the internet of things to developed efficient IDS (Da Costa et al., 2019, Hasan et al., 2019),
However less work has done to find out research trend in network security. So our contribution in this
paper is to find out the research gap from a large collection of abstract papers by using the topic modeling
approach and LDA technique of text mining. In this study, the authors have collected the data of 2000
abstract papers from the Scopus library. These collected data having the eight Columns such as Author’s
name, Title, Year of publication, Links of paper, Abstract, Author Keywords, Indexed keywords, and
References. After collecting authors are doing text cleaning process in which removing unnecessary
information which has less meaning in deciding the topic to further research gap, then after applying
20
Research Trends for Malware and Intrusion Detection on Network Systems
topic modeling approach to generate most frequent word occurring in the corpus of text data and finally
showing the data visualization of topics obtained from LDA approach to know the research gap.
LITERATURE REVIEW
Intrusion in any network is malicious code that moves across the network to monitor the network traffic
data over the communication channel among computers. It is an act of illegal activity which violates
the policy of digital information and tries to gain access into systems in an unauthorized way. There is
various form of malicious activity-travel across the computer network such as anomaly, malware, viruses,
worms, Trojan horses, etc. the intensity of damage varies according to the type of attack. there are various
areas where the spread of intrusion has been expanded and damages the systems critical resources, steal
private information, and violate confidentiality, integrity, and authentication such areas are the internet
of things, cloud computing, internet of vehicles, wireless sensor networks, big data systems and many
application of latest technology are untouched from various types of new attacks are growing. Therefore,
we need to protect the network resources as the utmost requirement of today’s time for various business
applications by employing different techniques of machine learning, deep learning, as well as data min-
ing tools also (Binbusayyis et al., 2020), brings into intrusion detection systems (IDS). In this paper, we
have shown related work of previous year papers of different areas from 2014 to 2021 related to intru-
sion detection, anomaly, and malware detection. This work is finding the research gap in the literature
by using topic modeling and recommended future directions of research in the areas of cybersecurity.
(Chhabra et al., 2013) provides the methods to control the DDoS attack in MANET and pointed out the
critical challenges of this attack that is difficult to prevent due to its dynamic nature of attacking. (Canfora
et al., 2014) discuss the intensity and impact of malware attacks and have proposed a new technique to
detect viruses, malicious activity from datasets by doing an in-depth analysis of changing patterns of
malicious code. Results of new techniques have shown better precision percentage in the detection of
malware code as well as classify their categories also.(R. Ray et al., 2014) have worked on identifying
inefficiencies in the areas of IoT that are prone to the security of devices such as privacy concerns and
authentication of various devices. In this paper, the author suggested the deployment of RFID in IoT to
provide a secure platform as well as the capability of detecting malware attacks. The result of the proposed
method achieved good security from the existing protocol and provides scalability to the IoT framework.
(Elhag et al., 2015) have proposed a model to improve detection accuracy by incorporating genetic fuzzy
21
Research Trends for Malware and Intrusion Detection on Network Systems
techniques inside the pairing-based learning model. This framework is based on two concepts, first is to
define the boundary label of sets and make the two classes of data set by using the divide and conquer
approach. (Jamali & Jafarjadeh, 2015) worked on the design of an intelligent automatic smart detection
system to detect attacks having no predefined structure by using the automatic learning approach. This
model provides satisfactory performance on KDD99 datasets from an existing system. (Feizollah et al.,
2016) studied the weakness of the Android system through which malware attack is major threat such
loopholes as openness of IPC with reusability feature provides the gap to attack and did an extensive
analysis of both types of data infected and clean data. (Anusha & Sathiyamoorthy, 2016) have given the
importance of selecting optimal features to increase the better detection accuracy for this used the optimal
feature selection techniques which is particle swarm optimization to remove the unnecessary features
and analyze the dynamic features for a new attack. (Wang et al., 2017) contribute in this work to apply
the ML classifier to the large collection of android applications to categorize them into the attack-free
app and infected with malicious activity. It has good results regarding the identification of a good and
bad category of apps. (Lee et al., 2017) have focused on the improving performance of machine learning
classification techniques by selecting the best subset of features of a dataset. For this author used two
approaches like sequential forward search to obtain reduced dimension and the second is random forest
classifier to achieve better classification accuracy.
(Gupta & Gupta, 2017) have discussed the security challenges, issues attacks, mitigation approaches
and various types of loopholes in web application to keep secure web application in era of digital age
from various cyber threat. (Zhang et al., 2017) proposed a prototype framework to protect the piracy,
copyrights issues in multimedia social networks named CyVOD. This framework prevents the digital
contents from transmitting in unauthorized way. (Jerlina & Marimuthu, 2018) explained the issues ac-
companied in previous work related to malicious activity in network systems such as more detection
time and computation time of unwanted activity in the network. It has removed the above concern by
applying feature extraction for preprocessing and generating the rule-based system by using the Rete
algorithm. (Vijayanand et al., 2018) also discussed factors of intrusion detection systems that may de-
grade the performance of the models such as irrelevant variables and redundant features for wireless
mesh networks. These issues have been overcome in this research work by support vector machines and
genetic algorithms. (Hong et al., 2019) has proposed a new concept of smart electronic device enabled
with intelligent detection system able to identify the behavior of data such as anomalous data and normal
data. This new concept depended on the nearest node to take rightful and accurate decisions in finding
the source of the attack on the embedded system. (Dwivedi et al., 2019) proposed a new approach of
increasing the security of network domain from various unwanted malicious code. This technique ap-
plied the group classifier to select the best attribute from multiple variables. The grasshopper optimizing
algorithm selects the best optimal subset of features to improve the detection rate of anomaly. (Hassan et
al., 2020) proposed a hybrid model of deep learning to extract hidden features of intrusion details from
large volumes of data called big data clouds. This deep learning approach is able to detect the unknown
structure of code through the technique of CNN. (Almogren, 2020) has explored the new potential threat
in the edge of things via relocation of data storage at edge nodes to efficiently processing of data. This
environment of edge nodes also having security issues so it could be overcome by a deep belief network
(DBN). It is a very effective and advanced approach to deep learning from the existing method about the
performance of intrusion detection systems. (Tewari & Gupta, 2020) have proposed a protocol mutual
authentication to safe the RFID tags from cyberattacks for IoT devices. This prevent the attack from using
the recurring rotation based approach on the authentication features. (Stergiou et al., 2020) worked on
22
Research Trends for Malware and Intrusion Detection on Network Systems
the 6G wireless technology to provide secure and innovative platform based on IoT for communication
between smart building and in fog network to handles the big data. This work support the secure brows-
ing and data sharing for organization and users in fog networks. (Guo et al., 2021) worked in the areas
of mobile application where authors have shown the concern of event call back issues not paid atten-
tion efficiently. This research work has solved the above issue by proposing call back based multilayer
embedding method as callback2vec. This new method categorizes the normal API and ordinarily API
call-backs. (Li et al., 2021) have pointed out the updating of ML classifiers to new types of attacks on
the internet of vehicles (IoV). It proposed a dual model-based updating approach to tackle unlabeled
data in the classifier using the transfer learning framework. It is time-efficient as well as achieving a
good accuracy level. (Kumar et al., 2021) proposed an intelligent cyber-attack detection system for IoT
network using the hybrid feature reduced approach and overcome the curse dimensionality problems.
The feature are combined using the AND operation to get the optimize attributes.
The figure 1 demonstrates that this work is done in four steps such as data collection, data cleaning,
topic modeling, and finally visualization of data. The process is broadly classified into two subparts that
are pre-processing of collected data and topic generation. The first part includes removing unnecessary
information, converting data into lower case, removing special symbols, punctuations, numbers, comma,
spaces, brackets, etc. To obtain a structured store of words (document term matrix) and useful pattern of
the word cloud is known as document term matrix for deciding topics frequency and interpretation of
topics. In the second stage, the authors would get the topics by applying LDA (latent Dirichlet allocation)
to generate the topics (Altaweel et al., 2019), and get important insights about the further research gap.
An unsupervised machine learning approach known as the Latent Dirichlet allocation (LDA) algorithm
is used for topic modeling from text data. It is a statistical probabilistic generative model to extract
hidden topics from unobserved documents. In-text data consist of a number of the document in which
document each document having the number of topics, further each topic contains the number of words
that may be similar or different. The LDA works based on a three-layer Bayesian probability calculation
model that provides the relative probability of each word. The Topics are defined as the distribution of
words present in the corpus of data. This distribution of topics may be in symmetrical or asymmetrical
order in documents. The LDA model is represented through a plate notation, which has some dependen-
cies variables. The LDA model has two boxes called plates as inner plate and outer plate; these plates
contained the repeated objects. The outer plates indicate the document and the inner plate contained the
repeated words from the available documents (Blei et al., 2003, Gangadharan et al., 2020, Wang et al.,
2018). The description of the variable used in the LDA model is defined as follows:
D is the number of documents, E is the number of words where document u has EU words, α is the
parameter of Dirichlet prior on per-document topic distributions, β represents the parameter of Dirichlet
23
Research Trends for Malware and Intrusion Detection on Network Systems
prior on the per-topic distribution, θs is the topic distribution for document u, λ is the word distribution for
topic T, ZU, V is the topic for a Vth word in document u, Wu, v is the specific words (Altaweel et al., 2019).
In figure 2, the observable variables are represented in the form of WU, V by shadow circle while
another circle indicates that the unstructured and unobservable variables (Altaweel et al., 2019). In this
LDA process, the original document is decomposing into the document-word matrix which represents
the corpus. The θ entity in the model shows that row having document and columns defined by topics,
whereas λ consists of rows described by topics and columns described by words. The general structure
of work has shown in figure 3.
DATA GATHERING
The data collection for study is done from the Scopus library which is a citation index database. A total
of 2000 research papers abstract have been collected related to network attacks through indexed key-
word, author keyword search. These papers are of 2014 to the 2021 year published in a various reputed
journals like Elsevier, Springer, IEEE computer society, information sciences, network, and computer
application, IEEE Transaction, etc.
PRE-PROCESSING
The author’s data is in structure form and having the eight columns like “Author”, “Title”, “Year”, “link”,
“Abstract”, “Author keywords”, “Indexed keywords” and “References”. The whole study of data collec-
tion is based on abstract collection from which find out the hidden topics for topic generation and convert
these abstract data into CSV (comma-separated values) file and import it into the R software, which is a
statistical programming tool. After converting data into CSV file format we start it’s cleaning and before
doing cleaning ensures that R software contains the required package such as “tm” and “topic models”.
24
Research Trends for Malware and Intrusion Detection on Network Systems
The cleaning process starts with the selection of abstracts from data and converts it into lower case and
removes the numbers, full stop, punctuations, and whitespaces and stop words. Stop words are those
which do not have any special meaning in deciding topics from the text(e.g., demonstrative pronouns
“this”, “these”, “that”, and “those”; indefinite articles “a(n)” and the definite article “the”; verb “to
be”, which causes changes in grammatical person and tense; and auxiliary verbs, such as “have”, “has”,
“do”, and “does”, that exist in the grammatical system without specific meanings). Once authors remove
these words, white space is created so authors have to remove that also, some of the words which are
identical but since their ending is different due to change in tenses and number(singular/plural) software
perceive it differently, so to remove it use the “stemming” operation, Where only the roots of words
are extracted. The tm package, a natural language (English) processing program within R, provides the
stemming function i.e. word “passed” and “passing” are the past and present progressive forms of the
verb “pass” so all these words present in the article are converted into its root word “pass” and counted.
After doing all of this operation get the collection of 7604 words. These words can be represented in
various ways; figure 4 is showing the word cloud which builds with the help of an R software package
called “wordcloud2” in this cloud words are shown with respect to their frequency. Basically, the word
cloud is used to represent the summary of cleaned data in visualize of various keywords in an oval shape,
circle shape, etc. the size of the keyword in the word cloud varies according to the term frequency of
word occurrence as demonstrated in figure 5.
TOPIC MODELLING
Topic modelling is an approach of text mining in which text is organize, structured and summarizes in
proper way to find out the hidden topics from large collection of data. It is an unsupervised machine
learning technique that has the ability to scan set of multiple documents, detect words, and symmetrical
expression of words present in corpus of data. Topic modelling can be used to classify the set of text
data into a few topics based on the frequency of occurrence of words in data.
25
Research Trends for Malware and Intrusion Detection on Network Systems
Topic modelling in this study uses a LDA package of R language which is a statistical model to
calculate probability (>0.5) of appearance of topics belongs to the particular topics. LDA convert all
the word set data into document term matrix then after calculate the sum of each row (document) and
eliminate empty documents. After this we need to make sample of whole text data by using Gibbs
sampling. Topic modelling on collected sample data, probability of appearance of specific words in
each topics, the distribution of topics in each paper by taking K=25(number of topics) initially, but
authors see in result that topics has many common words, so authors reduce the number of topics to
20 etc. after many iterations reach a conclusion that 15 topics are appropriate for our study, in figure 6
shown the top 15 words from each topic with their corresponding frequency given. The table 2 shows
that the topic with the corresponding maximum and minimum probability of words. Probability of each
document corresponding to each topic is calculated and extracted. And the highest probability of each
document corresponding to their topic is fall in that topic. Table 2 shows that the range of probability of
26
Research Trends for Malware and Intrusion Detection on Network Systems
maximum and minimum of all 15 topics. The topics1 has maximum 0.348008 and minimum 0.145022
probabilities. It means that 14.50% chances that document 7 is related to topic1 and its topic name given
“strategies to detect intrusion”. The topic 2 has maximum probability of 0.370766 that means there are
37.07% chances that belongs to document 14. Similarly for other remaining 13 topics also belongs to
corresponding document number as shown in table 1.
Documents 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Topics 7 6 17 13 11 17 1 13 19 18 4 14 4 2 10
This study has been done to extract the topics that have been researched and published in previous
work in network security papers abstract. The topic name has been given on basis of search keyword
appears frequently in documents. The top 10 most frequent word has been found in each 15 topics as
shown in figure.6. These are categorises into fifteen topics named topic 1, topic 2, topic 3 etc. To giv-
ing the name for each topic the authors selected the 11-13 an abstract belongs to a specific topic and
shown the authors keyword of the abstracts in table 1. For example to name topic 1, following keywords
have been taken from some of the abstracts such as Computer crime (Sovilj D et al., 2020), Advanced
metering infrastructures (Budnarain P. et al., 200), Anomaly detection (Sanner S. et al., 2020), Optimal
strategies (Salmon G., Rao M. et al., 2020). According to this keyword it can be observe that it related
to cyber-attacks prevention method, so the topic 1 named as “strategies to detect intrusion”.
For topic 2 the indexed keyword such as Artificial intelligence (Zhang Z et al., 2021), Deep learning
(Cheng Y. et al., 2021), Signature-based approach (Gao Y et al., 2021), Digital storage (Nepal S et al.,
2021), Feature extraction (Zou Y et al., 2021), from these keywords it can be seen that it is about the
“AI and ML methods for intrusion detection” as shown in table 3.
27
Research Trends for Malware and Intrusion Detection on Network Systems
Table 2. Topic with the corresponding maximum and minimum probability of words
The topic 3 keywords includes Personal computing (Jeon S et al., 2020), Data recovery (Moon J et
al., 2020), Bioinformatics, Sequence Covering Similarity, Host-based intrusion detection, System calls
(MartÃnez-GarcÃa H.A. et al., 2020) from these words it can be concluded that it is related to “Bio-
inspired intrusion detection”.
The indexed keyword for topic 4 as follows Future research directions (Khasawneh et al. 2020), Indi-
vidual classifiers (Ozsoy M et al., 2020), Different attacks (Donovick C et al., 2020), Ensemble learning
(Abu-Ghazaleh N et al., 2020), Insider attack (Ponomarev D. et al., 2020) according to it topic 4 named
as “Advance methods for intrusion detection”.
For the topic 5 the important words are Internet technology(Kaur S. et al., 2020), Intrusion Detection
Systems(Singh M et al.,2020), Logistic regressions(Shawly T. et al.,2020), Network intrusion detection
systems(Khayat M et al., 2020), NSL-KDD,XGBoost(Elghariani A et al., 2020), Network security(Ghafoor
A. et al., 2020) on the basis of these found words the topic 5 can be named as “Statistical methods for
intrusion detection”.
For topic 6 the search indexed key words are as Agricultural robots (Wang A et al., 2020), Denial-of-
service attack (Chang W et al., 2020), Home health care(Chen S. et al., 2020), False positive rates(Ieracitano
C et al., 2020), Industrial process(Adeel A et al.,2020), Intelligent Intrusion detection systems(Mohaisen
A., et al., 2020) from these words it can be seen that it is related to “Industrial applications on intrusion
detection systems” .
For the topic 7 the keyword from the documents are Digital libraries (Azarudhin et al., 2020), Dy-
namic analysis (Anithaashri et al., 2020), Intrusion detection (Patil N.M. et al., 2020), Learning systems
(Dias S.P. et al., 2020), Security systems (Dcunha A.A. et al., 2020), Virtual machine (Dodti R.J. et al.,
2020), Cloud securities, KVM introspection, Machine learning are indicating about the “Virtual data
security and applications”.
28
Research Trends for Malware and Intrusion Detection on Network Systems
The topic named given “Medical applications of Intrusion detection” for topic 8 on the basis of
found keywords such as Biochips (Aonzo S. et al., 2020), Diagnosis(Merlo A. et al., 2020), DNA
sequences(Migliardi M. et al, 2020), Drug delivery(Oneto L et al.,2020), Gene encoding(Palmieri
F. et al., 2020), Malware(Borah S. et al., 2020), Medical applications(Karsligil et al., 2020), Ap-
plication programs(Sun B et al., 2020), Evolutionary algorithms(Sharma J et al., 2020), Inspection
equipment(Yaacoubi O., 2020).
For the topic 9 keywords are includes Hamming distance (Taheri R et al., 2020), Malware (Ghahra-
mani M. et al., 2020), nearest neighbor search (Javidan R. et al., 2020), Static analysis (Shojafar M. et
al., 2020), All nearest neighbors (Pooranian Z. et al., 2020), Clustering (Conti M. et al., 2020), K nearest
neighbor (Romli R.N. et al., 2020). It is about detection techniques, so it is named as “Unsupervised
techniques for intrusion detection”.
The topic 10 keywords are Anomaly detection (Patil R. et al., 2020), Computer crime (Dudeja H. et
al., 2020), (Modi C. et al., 2020), Gradient methods (Pandey S.K. et al., 2019), Intrusion detection (Modi
C. et al., 2020), Momentum (Sekhar R. et al., 2019), Technical vulnerabilities (Krishnan P. et al., 2020).
These keywords are describing about cyber threat. So it named as “Security threats due to Intrusions”.
For the topic 11 the indexed keywords are Command and control systems (Ganeshan R et al., 2020),
Complex networks (Rodrigues P. et al., 2020), Computer crime (Thanudas B et al., 2019), Graphic
methods (Sreelal S. et al., 2019), Internet protocols (Cyril Raj V. et al., 2019), Internet service providers
(Maji S. et al., 2019), Botnet(Pao H.-K. et al., 2019); DNS, Domain names(Shi Y. et al., 2018), Graph
clustering(Chen G et al., 2018). This is explaining about research of network threat. So it is named as
“Network research on intrusion detection”.
For the topic 12 the keywords are Field programmable gate arrays (Abawajy J.H. et al., 2020), Intru-
sion detection (Chowdhury M et al., 2020), Learning systems (Kelarev A. et al., 2020), Benchmarking
(Mishra P. et al., 2020), Computer crime (Verma I. et al., 2020), Decision trees (Gupta S. et al., 2020)
from these words it can be named as “Firmware based intrusion detection systems”.
For the topic 13 keywords included are Contamination (Yang K. et al., 2018) Earth (Ren J. et al.,
2018), Infrared devices (Zhu Y. et al., 2018), Intrusion detection(Zhang W.,2018), Moon(Yang Y. et
al., 2018), Signal detection(Cai Z. et al., 2018), Sounding apparatus(Wang C. et al., 2018), Calibration
reference(Zhang J. et al., 2018,), Cross-track infrared sounder(Chen J. et al., 2018). These keywords are
showing some relationship towards environmental. So it is named as “Environmental applications on
intrusion detection”.
For the topic 14 there are some of keywords such as Chemical detection (Sahu S.K. et al., 2019),
Computer crime (Katiyar A. et al., 2019), Hidden Markov models (Kumari K.M. et al., 2019), Mal-
ware (Kumar G. et al., 2019), Markov processes (Choi W. et al., 2018), Mercury (Joo K. et al., 2018),
Commerce (Jo H.J. et al., 2018), Competition(Park M.C. et al., 2018), Computer crime(Lee D.H. et al.,
2018). These words are showing relationship about substance related detection. So this topic named as
“Intrusion detection systems based on substance identification”
For the last topic 15 the Computation theory (Guo Y. et al., 2019), Computer games(Zhang H. et al.,
2019), Game theory(Zhang L. et al., 2019), Trusted computing(Fang L. et al., 2019), Coalitional game(Li
F et al., 2109), Cooperative intrusion detection(Bhandari S. et al., 2019), Detection accuracy(Panihar
R. et al., 2019), Fairness assurance(Naval S. et al., 2018), Security(Laxmi V. et al., 2018), Stackelberg
Games(Zemmari A. et al., 2018). These are the word describing about the games related work. So this
is named as “Gaming applications and intrusion detection”.
29
Research Trends for Malware and Intrusion Detection on Network Systems
RESULT ANALYSIS
In this section, authors have obtained five tables in this study; table 1 contains the “document to topics”
that combines the number of documents into topics on basis of search keywords. The table 2 is cre-
ated by calculated corresponding maximum and minimum probability of fifteen topics using the LDA
technique, each topic showing relative maximum and minimum probability of words for example topic
1 has maximum probability 0.348008 and minimum probability is 0.145022 and likewise for remaining
topics, Maximum and minimum words frequency of top 10 words for 15 topics to find out the research
trends. Table 3 shows that topic wise some of the major keywords and citations, authors found fifteen
topics that get after applying the LDA algorithm, need to determine the name of topics based on words
frequency of appearance was considered first. And based on that write the name for each topic in table
3 with citations shows like in topic one has the index keywords “Advanced metering infrastructures,
Computer crime, Learning algorithms, Support vector machines, Training and testing, Anomaly detec-
tion, Optimal strategies” so authors write it as the “strategies to detect intrusion”. For topic 2 the indeed
keywords are “Artificial intelligence, Deep learning, Signature-based approach, Digital storage, Data
mining, Feature extraction, Pattern mining, Malware”. So on the basis of indexed keyword named it as
“AI and ML methods for intrusion detection” Similarly for others topic name also.
The table 4 shows that number of articles per topic, after applying the LDA authors get to know that
which document fall in which topic, and based on that we classify the entire document corresponding
to 15 topics, shown in table 4 it is seen in table that some of the topic has a greater number of document
while other are very less. This can be represented in better way by the graphical means shown in figure
7, there is in topic 15 the least no of document (57) are fall, it also mean that the research in this field
is not done very frequently as in other topics, so by this authors can conclude that there is a gap in this
field and can be explore more for research work. So finally authors came to know to that in field of the
network research on intrusion detection (topic11) research has been done least. Topic 2 has less number
of documents (76) comes, so in this topic also less research has been done. Similarly for topic 15 come
out 84 documents under this topic so authors can say that it is also less research done topic.
Table 5 in this study shows that the year wise article distribution of data of fifteen topics from 2013-
2014 up to 2019-2020 which indicates the research done of topics. In topic 1 that is related to “strategies
to detect intrusion” the article distribution of research growing up from 2013 to 2016 and from 2016
onwards to 2018 falls down. Again it is seen that a little rise in research from 2018 to 2020 has done.
For topic 2 “ AI and ML methods for intrusion detection” the article distribution among 2013-2014
and 2019-2020 indicating a continue rise up to 2015-2016, after 2016 has been observed the slowdown
in number of article from the graph till 2018 and again rise in article distribution from 2018 to 2020.
The article distribution for topic 3 that is related to “bio-inspired intrusion detection” in the year 2013
to 2016 was continuous rise around 35% then after seen constant distribution for bio-inspired intrusion
detection. The topic 4 is related to “advances method for intrusion detection” for this around 36% in-
crease in article distribution in the year 2013-2014 to 2015-2016 seen and later on it has been constant
contribution of article have seen till 2019-2020. For topic 5 from 2013-2014 to 2015-2016 the statistical
methods for intrusion detection has risen around 70% contribution of articles and then after it is seen
constant contribution from the curve from 2017-2018 to 2019-2020.
For the topic 6 which is related to “industrial application on intrusion detection system” the curve
showing very slowly increase in article distribution for this areas from 2013-2014 to 2019-2020 around
85% articles.
30
Research Trends for Malware and Intrusion Detection on Network Systems
For the topic 7 that is “virtual data security and application” the research trends for this in starting
has been growing slowly around 33% contribution of research articles observed.
The topic 8 which is indicating “medical applications of intrusion detection” also showing constant
rise of curve from 2013-2014 to 2019-2020 regarding the research trends in this field. The research article
trends for topic 9 which is related to “unsupervised techniques for intrusion detection” indicating sud-
den increase in articles from 2013-2014 to 2015-2016. In between 2016 and 2017 trend has decreased
slightly and then after it increases the from 2018 to2020. For the topic 10 that is “security threat due to
intrusions” the research has increased from 2013-2014 to 2019-2020 around 51.10%. Topic 11 related
to “Network research on intrusion detection” for this the article distribution trend of research has risen
from 2013-2014 to 2019-2020 around 32.50% from the graph can be seen.
The topic 12 that is “Firmware based intrusion detection system” involves the around 69% growth
over the year from 2013-2014 to 2015-2016, in between it has decreased the growth of articles a little
bit again from 2017-2018 to 2019-2020 the increment in research articles can be seen around 80%. For
the topic 13 that is “environmental application on intrusion detection” research has done around 25%
from the 2013-2014 to 2015-2016 and after that from 2017-2018 to 2019-2020 sudden growth in article
distribution up to 70% can be seen from the graph. In the topic 14 the research article distribution was
around 40% from 2013-2014 to 2019-2020 in the areas of “Intrusion detection system based on substance
identification”. For the last topic 15 related to “Gaming application and intrusion detection” showing
continuous growth over the years up to 2015-2016 from 2013-2014 of around 20% articles distribution
and slightly decrease in growth, again it can be seen rise up to 39% article distribution in this domain.
31
Research Trends for Malware and Intrusion Detection on Network Systems
Table 3. Continued
32
Research Trends for Malware and Intrusion Detection on Network Systems
DISCUSSION
In the context of intrusion and malware detection in security domain there are many things to consider,
like most important one is the security of users confidential data, integrity, authenticity as well as physi-
cal security also necessary to make the network system more trustable and robust from cyber-attacks.
Therefore to ensure the security of system model must have some necessary quality like intrusion
detection and prevention mechanism. So designing the good security framework model is a very high
33
Research Trends for Malware and Intrusion Detection on Network Systems
potential in growth of advancement in technology. Although a lot of intrusion detection and malware
detection system have been proposed by previous researchers but as the more use of internet in various
fields increase the threat of cyber-attack. so to understand the research trends in the field of cyber and
network security we here collect 2000 papers abstracts related to that and analyse through topic model-
ling to get insights about research gap in this area.
34
Research Trends for Malware and Intrusion Detection on Network Systems
Table 3. Continued
In the this study, the 15 topics which we get after multiple iteration regarding the intrusion detection,
according to perplexity calculation in topic modelling, the number of appropriate topics amounted to
15, all these 15 topics are related to different areas such as, Agricultural robots, Home health care, Bio-
informatics, Computation theory, Computer games, Game theory, Trusted computing, Vehicle security,
Graph clustering etc.
35
Research Trends for Malware and Intrusion Detection on Network Systems
Table 3. Continued
As all topics shown in the table 5 describes their keyword and topics names related to intrusion detec-
tion areas. The topic 1 shows that the strategies to intrusion detection and year wise article distribution
from year 2013 to 2020 in which around 52.4% research articles are based on the strategies of intrusion
detection. Topic 2 is related to AI and ML methods for intrusion detection which depicted that in year
from 2013 to 2016 around 43.30% AI and ML techniques applied and after that suddenly the scope of
these machine learning has been gradually increase to 49.6%. Topic 3 is about the Bio-inspired intrusion
detection; this shows that around 36.3% research related to these areas. Topic 4 and topic 5 are related to
36
Research Trends for Malware and Intrusion Detection on Network Systems
advance and statistical methods for intrusion detection in which around 36.7% and 61.8% respectively
work has been done. Topic 6, topic 7, topic 8, topic 13, topic 15 are based on the Industrial applications,
Virtual data security and applications, Medical applications of Intrusion detection, Environmental applica-
tions on intrusion detection, Gaming applications and intrusion detection respectively. This distribution
of articles in these fields is around 82.3%, 31.7% and 89.4, 70.3%, 35.8% research work respectively. The
percent wise distribution of remaining topics like topic 9, topic 10, topic 11, topic12, topic 14 research
trends is around between 31.4% to 41.8% related to Unsupervised techniques for intrusion detection,
Security threats due to Intrusions, Network research on intrusion detection, Firmware based intrusion
detection systems, Intrusion detection systems based on substance identification.
CONCLUSION
In the field of intrusion detection although many researches had been done over the years, but these
researches are not a very organised in a structured way. so it create a research gap over time, this study
aim to find research gaps and try to focus on that and bring it into the light. For execution of such task,
we use here a specific technique called topic modelling, in this technique through the function called
“LDA”. In this work, we got to know that there are some topic which are very less number of document
and research done. The graph in table 5, we observe the topic 15 is least no. of doc. So the research in
the field of Gaming applications and intrusion detection are not done in frequent ways as in other top-
ics, but analysing only one thing at a time not always give fruitful results. So here also we cannot just
relay on topic but also topic ten is important because security threat is one of the pillars of designing
the security framework to detect intrusion which lead towards more security parameters, good healthy
environment of network communication and lead to more transparent, secure, and reliable. In this study
although there are many limitations, some important are discussed in following section, one of them
major limitations are this study is based on analysis of the abstract and index keyword collected from
limited number of the article (2000). And considering that the sample (Gibbs sampling) which we use
for performing the “LDA” is representative of the whole populations. In addition on this study during
the topic name selection is based on some of the index keyword among many of there, so here also it
assume that these words correctly describe the topic name and the index keyword which taken from
some of the documents not from all which fall in that topic through the “LDA” i.e. in (topic 1) we have
total of 150 documents but the words chosen for the topic name are only from four documents, so it is
also a limitations of this study, it will be our future work to remove this limitations.
REFERENCES
37
Research Trends for Malware and Intrusion Detection on Network Systems
Anusha, K., & Sathiyamoorthy, E. (2016). A decision tree-based rule formation with combined PSO-
GA algorithm for intrusion detection system. International Journal of Internet Technology and Secured
Transactions, 6(3), 186–202. doi:10.1504/IJITST.2016.080399
Belavagi, M. C., & Muniyal, B. (2016). Performance evaluation of supervised machine learning algo-
rithms for intrusion detection. Procedia Computer Science, 89, 117–123. doi:10.1016/j.procs.2016.06.016
Binbusayyis, A., & Vaiyapuri, T. (2020). Comprehensive analysis and recommendation of feature
evaluation measures for intrusion detection. Heliyon, 6(7), e04262. doi:10.1016/j.heliyon.2020.e04262
PMID:32685709
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. The Journal of Machine
Learning Research, 3, 993-1022.
Canfora, G., Mercaldo, F., Visaggio, C. A., & Di Notte, P. (2014). Metamorphic malware detection using
code metrics. Information Security Journal: A Global Perspective, 23(3), 57-67.
Chhabra, M., Gupta, B., & Almomani, A. (2013). A novel solution to handle DDOS attack in MANET.
Academic Press.
da Costa, K. A., Papa, J. P., Lisboa, C. O., Munoz, R., & de Albuquerque, V. H. C. (2019). Internet of
Things: A survey on machine learning-based intrusion detection approaches. Computer Networks, 151,
147–157. doi:10.1016/j.comnet.2019.01.023
Dwivedi, S., Vardhan, M., Tripathi, S., & Shukla, A. K. (2020). Implementation of adaptive scheme in
evolutionary technique for anomaly-based intrusion detection. Evolutionary Intelligence, 13(1), 103–117.
doi:10.100712065-019-00293-8
Elhag, S., Fernández, A., Bawakid, A., Alshomrani, S., & Herrera, F. (2015). On the combination of
genetic fuzzy systems and pairwise learning for improving detection rates on intrusion detection systems.
Expert Systems with Applications, 42(1), 193–202. doi:10.1016/j.eswa.2014.08.002
Farnaaz, N., & Jabbar, M. A. (2016). Random forest modeling for network intrusion detection system.
Procedia Computer Science, 89, 213–217. doi:10.1016/j.procs.2016.06.047
Feizollah, A., Anuar, N. B., Salleh, R., Suarez-Tangil, G., & Furnell, S. (2017). Androdialysis: Analysis
of android intent effectiveness in malware detection. Computers & Security, 65, 121-134.
Gangadharan, V., & Gupta, D. (2020). Recognizing Named Entities in Agriculture Documents using
LDA based Topic Modelling Techniques. Procedia Computer Science, 171, 1337–1345. doi:10.1016/j.
procs.2020.04.143
Gibert, D., Mateu, C., & Planes, J. (2020). The rise of machine learning for detection and classification
of malware: Research developments, trends and challenges. Journal of Network and Computer Applica-
tions, 153, 102526. doi:10.1016/j.jnca.2019.102526
Guo, C., Huang, D., Dong, N., Zhang, J., & Xu, J. (2021). Callback2Vec: Callback-aware hierarchical
embedding for mobile application. Information Sciences, 542, 131–155. doi:10.1016/j.ins.2020.06.058
38
Research Trends for Malware and Intrusion Detection on Network Systems
Gupta, S., & Gupta, B. B. (2017). Detection, avoidance, and attack pattern mechanisms in modern web
application vulnerabilities: Present and future challenges. International Journal of Cloud Applications
and Computing, 7(3), 1–43. doi:10.4018/IJCAC.2017070101
Hasan, M., Islam, M. M., Zarif, M. I. I., & Hashem, M. M. A. (2019). Attack and anomaly detection in
IoT sensors in IoT sites using machine learning approaches. Internet of Things, 7, 100059. doi:10.1016/j.
iot.2019.100059
Hassan, M. M., Gumaei, A., Alsanad, A., Alrubaian, M., & Fortino, G. (2020). A hybrid deep learning
model for efficient intrusion detection in big data environment. Information Sciences, 513, 386–396.
doi:10.1016/j.ins.2019.10.069
Hong, J., & Liu, C. C. (2017). Intelligent electronic devices with collaborative intrusion detection sys-
tems. IEEE Transactions on Smart Grid, 10(1), 271–281. doi:10.1109/TSG.2017.2737826
Jamali, S., & Jafarzadeh, P. (2017). An intelligent intrusion detection system by using hierarchically
structured learning automata. Neural Computing & Applications, 28(5), 1001–1008. doi:10.100700521-
015-2116-4
Jerlin, M. A., & Marimuthu, K. (2018). A new malware detection system using machine learning tech-
niques for API call sequences. Journal of Applied Security Research, 13(1), 45–62. doi:10.1080/1936
1610.2018.1387734
Kumar, P., Gupta, G. P., & Tripathi, R. (2021). Toward Design of an Intelligent Cyber Attack Detection
System using Hybrid Feature Reduced Approach for IoT Networks. Arabian Journal for Science and
Engineering, 46(4), 3749–3778. doi:10.100713369-020-05181-3
Lee, J., Park, D., & Lee, C. (2017). Feature selection algorithm for intrusions detection system using
sequential forward search and random forest classifier. Transactions on Internet and Information Systems
(Seoul), 11(10), 5132–5148.
Li, X., Hu, Z., Xu, M., Wang, Y., & Ma, J. (2021). Transfer learning based intrusion detection scheme
for Internet of vehicles. Information Sciences, 547, 119–135. doi:10.1016/j.ins.2020.05.130
Ray, B. R., Abawajy, J., & Chowdhury, M. (2014). Scalable RFID security framework and protocol sup-
porting Internet of Things. Computer Networks, 67, 89–103. doi:10.1016/j.comnet.2014.03.023
Ren, Z., Chen, G., & Lu, W. (2020). Malware visualization methods based on deep convolution neural
networks. Multimedia Tools and Applications, 79(15), 10975–10993. doi:10.100711042-019-08310-9
Stein, G., Chen, B., Wu, A. S., & Hua, K. A. (2005, March). Decision tree classifier for network intru-
sion detection with GA-based feature selection. In Proceedings of the 43rd annual Southeast regional
conference-Volume 2 (pp. 136-141). 10.1145/1167253.1167288
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based big data secure management in the
fog over a 6G wireless network. IEEE Internet of Things Journal, 8(7), 5164–5171. doi:10.1109/
JIOT.2020.3033131
39
Research Trends for Malware and Intrusion Detection on Network Systems
Tewari, A., & Gupta, B. B. (2020). Secure Timestamp-Based Mutual Authentication Protocol for IoT
Devices Using RFID Tags. International Journal on Semantic Web and Information Systems, 16(3),
20–34. doi:10.4018/IJSWIS.2020070102
Vijayanand, R., Devaraj, D., & Kannapiran, B. (2018). Intrusion detection system for wireless mesh
network using multiple support vector machine classifiers with genetic-algorithm-based feature selec-
tion. Computers & Security, 77, 304–314. doi:10.1016/j.cose.2018.04.010
Wang, W., Feng, Y., & Dai, W. (2018). Topic analysis of online reviews for two competitive products
using latent Dirichlet allocation. Electronic Commerce Research and Applications, 29, 142–156.
doi:10.1016/j.elerap.2018.04.003
Wang, W., Li, Y., Wang, X., Liu, J., & Zhang, X. (2018). Detecting Android malicious apps and catego-
rizing benign apps with ensemble of classifiers. Future Generation Computer Systems, 78, 987–994.
doi:10.1016/j.future.2017.01.019
Zhang, Z., Sun, R., Zhao, C., Wang, J., Chang, C. K., & Gupta, B. B. (2017). CyVOD: A novel trin-
ity multimedia social network scheme. Multimedia Tools and Applications, 76(18), 18513–18529.
doi:10.100711042-016-4162-z
40
41
Chapter 3
Deep-Learning and Machine-
Learning-Based Techniques
for Malware Detection and
Data-Driven Network Security
Praneeth Gunti
National Institute of Technology, Kurukshetra, India
Brij B. Gupta
National Institute of Technology, Kurukshetra, India
ABSTRACT
A never-ending fight is taking place among malware creators and security experts as the advances
in malware are daunting. The machine learning strategies are indeed the new mode of researching
malware. The purpose of this chapter is to explore machine learning methods for malware recognition
and in general deep learning methods. The chapter gives complete explanations of the techniques and
resources used in a standard machine learning process for detecting malware. It examines the study
issues that are posed by existing study methods and introduces the potential avenues of study in future.
By administering a study to the participants, scholars have a better knowledge of the malware detec-
tions. The authors start by discussing simple dynamic modelling methods, their importance to the data
analytics of malware, and their implementations. They use open access resources such as virustotal.com
that review sample of dynamic analysis in reality.
DOI: 10.4018/978-1-7998-7789-9.ch003
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
INTRODUCTION
Throughout this section, you will understand the fundamentals of complex malware detection. Unlike
static research, which relies on what kind of malware seems like in document format, the dynamic analy-
sis comprises executing malware in a secure, confined area and then determine how it acts. It would
be like putting a harmful bacterial strain over to an enclosed area to see the impact against many cells.
Utilising dynamic analysis, we may switch through typical static analysis obstacles, like packaging and
deception, and also obtain more precise visibility into the intent of a specific malware test. Security and
Privacy is needed everywhere like in web applications (Gupta 2016), IoT devices (Tewari, 2017)(Ab
Malek, 2016), cloud computing (Al-Qerem, 2020), routing protocols (Jerbi, 2020) in WSN etc., so that
our data and the transactions are protected.
To grasp why dynamic analysis need, remember the topic of packaged malware. Recognise that pack-
aging malware relates to compacting or misrepresenting the x86 base source code of the software to
conceal the deceptive existence of the application. The bundled malicious code unwraps its own as it
attacks the target computer of that kind by which the program could be executed. We can attempt to
dismantle a sealed or abstracted malicious payload utilising static detection methods; however, it is a
tedious method. For instance, through static testing, we would first have to identify the position of the
ambiguous script within the malicious document. After that, we will need to locate the deception func-
tions that will successfully de-obfuscate this script to be executed. After finding the macros, we would
discover why this de-obfuscation technique operates to execute that on the program. And afterwards,
will we start the actual method of reversing the malicious script. An easy but intelligent solution to this
method is running the malicious program in a secure, enclosed atmosphere termed a sandbox. Testing
malicious code in a sandbox helps this one disassembles its own as it does before harming a primary
target. By merely executing malware, we will figure out how many servers a certain binary malware
binds to, which machine configuration settings it adjusts, as well as which Input/Output interface it is
attempting to do.
Dynamic modelling is valuable for both malware processing and malware reverse engineering. Since
dynamic modelling shows how a malicious test performs, we may equate its behaviour with many certain
malware mixtures. For instance, since dynamic analysis reveals which documents the malware suspects
send to the filesystem, we could use this information to link the malicious documents that write related
config files to the disc. These kinds of hints allow one to categorise malicious files focused on general
characteristics. It will also support us in recognising malicious files created by similar organisations or
are members of common groups. Most notably, dynamic processing is valuable for creating machine-
based malware indicators. We may prepare a detector to differentiate between harmful and benevolent
clones by analysing their actions through dynamic study. For instance, by analysing millions of complex
review reports from either malware or innocuous archives, a machine learning algorithm will understand
that if msword.exe executes a program called powershell.exe, this behaviour is harmful; however, while
msword.exe runs Browser, it is pretty safe.
42
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
You will access various downloadable, accessible, dynamic analysis resources online—this segment
reports on virustotal.com and CuckooBox. The virustotal.com platform has a web application that lets
anyone apply complex review samples for free. CuckooBox is an application framework that helps you
configure your custom dynamic testing system such that you could always test binaries locally. The
developers and administrators of virustoral.com even manage CuckooBox under the hood. Therefore,
knowing how to interpret the data on virustotal.com would help one analyse the outcomes of CuckooBox.
Below were the main groups a malware can undertake and the threat forms.
• Editing the file system, for instance, copying a machine driver to memory, updating machine con-
fig files, inserting new programmes to the operating system, and adjusting the windows registry
to guarantee start-up software.
• Editing a Windows registry for changing the device config. For instance, adjust the firewall
configurations.
• Setup application driver, for instance, processing a system driver to monitor consumer keyboard
strokes.
• Actions on the network, for instance, resolve web addresses and render HTTP queries.
We would discuss these habits in more depth utilising malicious payload and review the virustotal.
com analysis.
Dynamic analysis is an effective technique, but it’s not a solution for malware assessment. It has signifi-
cant restrictions. One drawback is that malware developers are conscious of CuckooBox as well as other
hierarchical testing systems and aim to bypass these by having their malware unable to function as it
notices that it is operating in CuckooBox. The CuckooBox engineers are conscious that malware writers
are attempting to do just that, but they try to get through tries by malware and bypass CuckooBox. This
cat and mouse activity is played on an ongoing basis so that some malware tests can eventually recognise
that they will be operating in dynamic analysis conditions and would struggle to function as we attempt to
execute them. Another drawback is that dynamic detection could not expose critical malware behaviours
despite any effort to bypass them. Take the problem of discrete malware which links directly to a remote
server while operating and waiting for instructions to be given. For e.g., such instructions can request
a malware template to scan for some types of target hosting data, record keyboard strokes, or switch
on a camera. Throughout this scenario, if the remote server does not transmit instructions when it is no
longer online, neither of such harmful actions can be observed. Owing to these drawbacks, dynamic
processing is not a cure for malware analysis. In reality, skilled malware researchers mix dynamic and
static research to deliver the optimal results feasibly.
43
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
REVIEW OF MALWARE
The method of malware deconstructing to learn how it operates, evaluate its features, root and possible
effect is named malware analysis. With millions of new malicious programmes in the wild and mutated
copies of previously identified programmes, the cumulative amount of malware identified by protec-
tion researchers has increased in recent years. As a consequence, malware detection is essential to any
company or network that leads to data breaches.
XSS vulnerabilities can be exploited by adding JavaScript function calls and function description
to perform harmful acts. (Gupta, 2016)(Gangwar, 2015)(Gupta, 2015) Address this by implementing
server-side XSS attack prevention and detection solutions, called XSS-SAFE, focused on automatic
injection information and disinfectant insertion into the JavaScript injection code.
The ad hoc mobile network is a less networked service owing to its operational capacity, without
any existing infrastructure being supported. In MANET, safety plays a crucial role because of applica-
tions such as battlefield and disaster relief networks. The absence of trustworthy unified authority and
scarce infrastructure means that MANETs are more insecure than wired networks. A system for handling
DDoS attacks in the mobile ad hoc network is urgently needed. (Chhabra, 2013) addressed the differ-
ent mechanisms for attacking the threats and the problems caused by DDoS attacks, including how the
attacks can influence MANET.
(Zhang, 2017) suggested a comprehensive multimedia networking site trinity method, designed the
algorithm for an enhanced hybrid suggestion by combining interactive filters and content-based recom-
mendations and created the CyVOD working prototype. Mobile multimedia DRM modes are added to
the network, including offline and online. Managing digital ownership is accomplished by digital content
encryption and secured access rights. In addition, security protocols are introduced that validate user
identification to prevent unauthorised users from transmitting their digital content maliciously. Compared
with the only content-driven approaches, the enhanced hybrid proposed methodology used in CyVOD
has upgraded the performance.
In particular, the researchers used the Attack Evaluation Practice, which describes a number of aspects
to help estimate the extent of danger related to known vulnerabilities. Ten of the various vulnerabilities
protected by (Gupta, 2017) was based on clear vulnerabilities.
There have been two main methods of malware detection: static analysis and dynamic analysis
(Gibert, 2020). On the one side, the static analysis requires testing the malware without executing it.
The dynamic analysis, on the other hand, includes executing the malware. Segments 2.1 and 1.2 offer
a brief overview of all methods.
Static Analysis
The static analysis comprises of analysing the program or configuration of the executable without load-
ing it. This review method will validate if a document is malicious, including usability details and can
even be often used to make a clear collection of signatures. For e.g., the most popular technique used
to recognise malicious software uniquely is hashing. The hashing software generates a special hash, a
kind of fingerprint that distinguishes the programme. The two main standard block ciphers are Message-
Digest Algorithm 5 (MD5) and Secure Hash Algorithm 1 (SHA-1). The most popular method to static
evaluation is:
44
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
• Locate sets of plot lines or strings. Trying to search via the sequences of a programme is the best
way to provide clues regarding its features. Binary derived sequences can include connections to
document locations of files changed or executed by the application, URLs viewed by the software,
domain names, Email addresses, threat instructions, dynamic link libraries (DLLs) identities en-
abled registry entries, etc. The Strings search method may be used to check for ASCII or Unicode
strings that neglect meaning and encoding in the runtime.
• Collecting the related resources and features of an application, and also the documentation re-
garding the file. These numbers include a census of which code is exchanged among software
solutions. The titles of these Windows features help one grasp what they are doing. The Utility
Dependency Walker is a free Windows framework that used a list of borrowed and transferred
features from an executable.
• PE headers and parts review. The import route headers contain more detail than only the last im-
port file names. The files included metadata on their own, like the parts of the real files. To obtain
this knowledge, you must use the PEView method.
• Looking for shrouded or encrypted code. Malware authors utilise packaging and coding techniques
to render their data impossible to examine. Programs packaged or encrypted provide shorter and
greater entropy than legal programmes. To identify loaded files, PEiD is used.
• Unplugging the programme or interpreting computer code in a programming language, i.e. de-
compiling the application. Some reverse-engineering methods load executables into disassem-
blers to explore what the programmes do. In this case, the essential resources for unplugging
Windows PE format executable are Ghidra, Radare2 and IDA Pro.
Dynamic Analysis
Runtime Functional Analysis permits the operation of the software and viewing the actions of the pro-
gramme on the device. This method is usually conducted when dynamic analysis has hit a dead-end or
completed static analysis. Dynamic research looks for the individual operations undertaken by a ma-
chine. A successful external review must be run in a secure setting where the device is not at risk, and
all networked networks are safe. In this way, dedicated utilities are equipped.
The computer must be physically separated from the Internet or some other network to avoid malware
from spreading. There are also protection concerns affecting several of the networked physical equipment.
The second possibility is to use “data computers” to do transient analysis. “Virtual system” is like a
device with some functionalities. The OS operating on the VM is maintained separated from the host
OS, and hence, the malware can’t damage the host OS. VMware, Oracle VM and VirtualBox are the
virtual device instances implementations that are used. There is a range of methods on the market to do
complex research on machine code. The Cuckoo Sandbox is recognised as the best free software malware
analysis framework. This method helps researchers to track API calls, evaluate internet traffic, and run
storage evaluation. There are different surveillance methods to track illegal acts and seek out fraudulent
activity. Process Monitoring or procmon is a software for Windows which monitors machine or process
operation. Task Manager shows different details on what processes are installed into the operating system.
Regshot is a registry comparison registry that enables registrations to be recorded. NetCat is a network
monitoring application. Wireshark is a free software data packets analyser that can be used to maintain
track of internet traffic. Besides tech, debuggers are still an important weapon. A debugger may be used
to test the routines of other applications. They give an informative description of a strategy and how a
45
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
programme unfolded. The perfect platform for malware researchers is OllyDbg since it’s open and has
several modules to expand its capability.
It’s a challenge that persists against the usage of virtual machines and sandboxing in virus and mal-
ware detection. And if you undertake all care when downloading applications, there will still be a chance
of contamination when using them. From period to period, vulnerabilities are discovered in VMware
software frameworks that enable a malware threat to be exploited.
Malware Development
The danger faced by malicious apps is tremendous, and there are a number of complex, advanced and
easy-to-access malware to screw up critical networks. Malware is often continually changing and forced
to keep up with threat defence. Malware campaigns improved following the rise in the polymorphic and
metamorphic methods used to track, remove and conceal the existence of the malware. Polymorphic
malware transforms the program while retaining untouched features of the malware software. Encryp-
tion and packaging are techniques that are used to conceal code. Compression may be used to mask the
essence of an application or software. Finally, at runtime, data reconstruction is performed. Malware
operators build resources such as crypters to encrypt or modify malware code so that it is more chal-
lenging for scholars to analyse it. A crypter helps you to decode malicious data, which means that the
code is secure. The malware updates the code on re-distribution or copies itself. Malware writers may
employ several transforming strategies including, though not limited to, renaming, permuting, extend-
ing, shrinking and inserting garbage code into their malware. Because of the extent of malware, the
procedures required to study malware are sometimes lengthy and complex.
Orthodox antivirus approaches, which are signature-based and heuristic, are quite time-consuming.
A signature is an attribute that identifies an exe file, equivalent to a fingerprint. Signature-based tech-
niques are unsuitable for identifying unidentified malware variants. We should seek more behavioural
research tools that can easily spot ransomware. To solve some of the difficulties of conventional antivirus
programs and maintain up with emerging ransomware threats and disruptive behaviour shifts, security
researchers built and used machine learning strategies in their approaches.
During the last decade, there’s been a sharp growth in computer educational methods to cope with malware
issues. The progress of machine learning would never be feasible without these technological advances.
The progress of machine learning would never be feasible without these technological advances.
• The first improvement is the substantial rise in viruses branded and counted that has been pub-
lished to the security world as well as to the academic circle. The spectrum of these streams varies
from small high-quality large-scale samples to massive quantities of malware (Ronen, 2018).
• The second trend is that computing capacity has considerably improved and is getting cheaper.
This resulted in speeding up the training phase and adapting more powerful and more sophisti-
cated learning algorithms to the continued rapid growth of data.
• Machine learning has made several advancements over the past decade and has scaled on a wide
variety of activities.
46
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
A timeline is collecting data in data science, cleaning it, choosing the necessary models, training
specific models, and eventually conducting forecasting. See Fig. 1. Rather than handling the malware
like a virus on the kernel level, most data preparation methods include pre-processing the exe file to
retrieve a series of functions that include an intuitive overview of the programme. The sample data has
been utilised to teach a method to optimise a specific problem. It is necessary to identify and discriminate
between various forms of malware to identify the threats effectively. The key variation among machine
learning systems is the way it recovers security data. A single value y = f(x) is extracted by a malware
detection method that ranges from 0 to 1. On another side, a classification method produces the likeli-
hood of a provided executable contributing from each output variable or group.
The classification system of the characteristics is given in Fig 2. According to this discourse, the
forms of malware detection methods may be classified into two types: static aspects and dynamic aspects.
Standard machine learning methods use professional experience because of manual data collection. This
offers an analytical perspective of how machine learning tools provide access to knowledge regarding
malware-related behaviour. The primary facets of a good machine learning project were feature architec-
ture and function extraction. Along with recent developments in computer vision and natural language
processing (NLP) study, there has been an advent of machine learning methods for malware classifica-
tion. These methods used in combination with ML. have rendered workflows entirely trainable from raw
data to the production of known objects. These methods are also used in securing IoT devices and SDN
networks (Letteri, 2019), Two-Phase Load Balancing Algorithm for Cloud Environments (Singh, 2021).
47
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
There are several methods in the usage of deep learning for identifying, classifying malicious ap-
plications, adversarial attacks (Mani, 2021). The goal audiences and groups concerned are described
in depth below.
Machine learning approaches in this segment can deliver a collection of features that are a sequence of
code. Then, the function vector is forwarded to a flow forward neural network. The obtained feature
vectors can be used in training feed-forward neural networks as well.
(Saxe, 2015) also addressed implementing the neural network malware detection method that consists
of three key components. The scheme is constructed of the five components as below: Their method was
tested on a dataset of 431926 exe files and obtained a recognition accuracy of 95%.
(Huang and Stoke, 2016) indicated an algorithm for malware identification and recognition focused
on deep learning. They derived static and interactive experiments, such as null-terminated cards, activ-
ity plus variable, and trigrams. Due to the increased dimensional space of input space, unsupervised
dimensionality reduction was made to create descriptive functions to classify each class. After that, the
variable of the function was decreased to 5000 characteristics utilising Random Principal Component
Analysis (RPCA). Finally, a deep architecture was educated utilising predictive features.
48
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
Researchers have pointed out a malware categorisation framework that map a high-dimensional
function map into a significantly low dimensional using randomly created vectors. Random forecasts
decreased the number of features from 180,000 to 4000 (Dahl, 2013). After the network sort learnt a
non-linear paradigm to identify malware, The device was tested on a database of 2.6 million labelled
items and an output ratio of 0.49%.
Image-Based Representation
The input for neural network approaches contains picture representations of binary material. Instead
of focusing on hand-engineered attribute separators, they tried implementing a fully connected neural
network framework for object training and testing for greyscale pictures.
(Gibert, 2019) used a Convolutional Network Design consisting of three convolutional chains and
one connected block. Each convolutional chain consisted of ReLU activation, limit and widespread ac-
ceptance. The convolutional levels served as attention detectors for the particular attributes or trends
throughout the input data and the properly connected layers to integrate the observed functions and
calculate a specific output. People’s success in grouping was assessed towards handcrafted attribute
extraction methods (Nataraj, 2011)(Kancherla, 2013)(Ahmadi, 2016), and they performed well. In this
review, (Rezende, 2017) proposed to use the ResNet-50 design and pretrained weights for malware
picture classification (Nataraj, 2011).
The method used in API function calls is to create an input functionality vector, and the vector’s loca-
tion shows if the application triggered a specific API function. It’s doesn’t recognise the series in which
all the API operations might have been used. Another alternate solution is to record the series of API
function calls and use it to construct classifiers.
(Athiwaratkun, 2017) sought to reflect the connection relationships between API behaviour of various
sizes. They also worked with Long Short-Term Memory and Gated Recurrent Unit as a paradigm for
expression. Their strategy occupies two phases. In the first point, we shall build our function candidates
by using LSTM or GRU. Ses attributes are classed with regression analysis then softmax. They also came
up with a convolutional neural network (Zhang, 2015). This network takes as input a series of around
1014 chars, each character being an activity. Patterns shorter than 10,000 chars are padded out at the
edge of a series with 0 bytes. There are nine layers in the character-level structure, two fully connected
layers and seven convolutional layers.
(Kolosnjaji, 2016) examined whether we can use neural networks (NN) to enhance the grouping of
freshly obtained malware variants into predetermined malware groups. They tested two forms of neural
networks, convolutionary, for modelling request series. They developed a neural network architecture
that incorporates the regularisation of n-grams and an entirely sequential model. Also, the API call series
of vulnerabilities without API requests replicated greater than two times in a sequence. Each specific
call was defined by one-hot encryption to identify variable per each request. It requires a convolution
accompanied by combining that serves as the function extractor. Thus, the fingerprints in the convo-
lutional portion of the system rely on the persistent portion of the system. To compute an aggregate
average, a pooling method was used. They added dropout to avoid overfitting and a softmax layer for
performance probabilities.
49
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
Guidance Cases
Software may be modelled as a series of procedures. These guidelines can be derived from static evalu-
ation and dynamic research. On the other side, you may inquire into acquiring them by unplugging
the executable code. Security administrators may also track and map the series of orders performed at
runtime. These strings of guidelines may be utilised to teach an end-to-end device to learn the required
functions without needing to explicate a million examples during preparation.
(Gibert, 2017) proposes a Neural Network that takes an encoding as a function compared to a Con-
volutional Neural Network. Convolution operation will learn to recognise n-gram-like signs since it
recognises subsets of the script. It helps them identify quite long n-grams, which are inefficiently cal-
culated using a clear evaluation of all n-grams. This is done by computing for varying sizes of different
variables. For e.g., in their study, the researchers incorporated 64 filters of length h x k per each h Î {2,
3, 4, 5, 6, 7}. The highest amount was selected as the equivalent function of the filter (also known as
global max-pooling). By doing this, we will find basic terms from input records. The softmax framework
outputs the likelihood over each of the groups.
Another possible way to cope with a hierarchical system of PE exe files is to tackle it with a Hierar-
chical Convolutional Neural Network (HCNN) (Gibert, 2019). Instead of displaying malware as a linear
set of commands, the author described them according to the functions of machine programmes. As a
result, assembly language commands were broken into components with the aid of mnemonics. As a
result, the hierarchy of convolutional functions grasped both the mnemonic-level and the more realistic
function-level of functions.
Digital Representation
A computer programme is also described using sequences of octets. In other terms, each character is a
unit of an input string. The most significant benefit of utilising the graph is that it can reflect malware
indistinctly if it is a PE file or a non-PE (including an ELF file). This poses a major challenge in por-
traying an executable as a series of bytes. First, by considering each character as a unit in a series, the
series would consist of millions of measures, rendering it one of the most demanding string categorisa-
tion tasks. Each byte signifies unique knowledge. Many binary files display varying degrees of spatial
similarity. Machines prefer to execute orders close to one another, but this law does not often apply due
to system requests and hops, and so commands and operation may be passed to other references in the
cache. These differences are held on the differential file and its hexadecimal type, respectively. Such
considerations should be addressed when constructing a system of analysis that identifies malware from
a series of bytes.
Use Convolutional Neural Network (CNN) to classify object to generate object recognition instead
of object category recognition (Raff, 2017). They merged the convolutional triggering with a universal
maximal pooling stage before a completely linked layer. Instead of taking raw values as data, they mapped
each byte to a fixed-size function vector.
(Krcál, 2018) completed an information-rich design consisting of a hidden layer, accompanied by
four convolutions, four max-pooling layers, another four convolutions, preceded by mean international
pooling then four completely linked layers. They tested their design towards the MalConv framework
and experienced an improvement of the predictive efficiency of about 1%. Besides, they handcrafted
functionality to create a better classifier.
50
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
Network Traffic
This system was suggested by (Prasse, 2017)(Muna, 2018) to identify malware on user machines. The
researchers retrieved separate functionality for client machines and site names from the obtained net-
work streams from the client-side. Then LSTM or recurrent neural network takes streams of inflow as
feedback and trains to decide whether the streams derive from malicious programs.
(AL-Hawawreh, 2018) defined intrusion detection techniques focused on deep learning frameworks
to identify breaches in Network ICSs. A Deep Autoencoder knows regular network patterns in the
unsupervised learning process, and a Deep Neural Network utilises the approximate variables of the
Autoencoder to fine-tune its variables and identify connection requests predictions.
A novel, complex image recovery model driven by visual salinity is described by (Wang, 2020). Im-
age forgery detection approaches include certain pattern descriptives to identify the image and run ML
algorithms to detect whether or not the image is fabricated. In the method, a noise map with a Wiener-
filter-based noise-reducing procedure is retrieved from the picture. The noise map is implemented
with an interrelation with both a pixel and the adjacent pixels using a multi-resolution regression filter
(Ghoneim, 2018)(Yamaguchi, 2017).
Mostly with Free Software Machine Learning Resources accessible nowadays, you could create cus-
tomised, machine-based malware identification tools, either as the primary identification mechanism or
to support troubleshooting tools, with minimal difficulty. So why develop one’s own machine learning
software while professional antivirus approaches are always accessible? Once you get access to real-life
instance breaches, like ransomware used for a specific community of intruders attacking your system,
creating personal new machine learning-based identification technology will provide us with training
samples of such vulnerabilities (Saxe, 2018).
On the other hand, commercial virus protection machines can ignore these warnings unless they
have malware signatures. Business methodologies are indeed “locked books”—that means we may not
recognise how they function and have a local opportunity to adjust it. While we develop our detector
systems, we know what they’ve done, and we can adapt it to the preference so that we minimise false
positivity and false negativity. This is useful as in certain implementations, and you may be able to
accept more significant false positives in return for fewer false negatives (for instance, while you’re
scanning the network for malicious items such that you might hand-inspect them to decide whether they
were fraudulent). In contrast, you may be prepared to forgive more significant false negatives in certain
implementations to return besides less false positives.
Throughout this part, you will explore how to build your custom high-level detecting software. I
begin by illustrating the fresh thoughts underlying machine learning, involving storage capabilities,
judgement limits, testing set, inadequately and over-fitting. After which, we concentrate on four simple
methodologies— random forest, logistic regression, decision tree and nearest neighbours how they could
be used to diagnose.
51
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
There is indeed a significant differentiation among other programming methods and machine learning.
While the standard algorithms instruct the programme what to perform, machine learning programs learn
by experience how and when to solve problems. For example, instead of merely taking out a series of
preloaded guidelines, machine learning vulnerability monitoring mechanisms could be programmed to
decide if a data is harmful or safe through analysing through observations of safe and dangerous data.
The expectation of information protection machine learning techniques was how they could robotise
generating signatures and then have the competence to accomplish better reliably than signature-based
malware analysis strategies, particularly on modern, previously undiscovered malware. Primarily, the
procedure we adopt to create every machine-based detector, such as a decision tree, comes out to the
following points:
• Set of malicious and benignware cases. We can use such scenarios (named learning scenarios) to
teach the framework of machine learning to recognise vulnerabilities.
• Remove the characteristics within each learning scenario to display the method as a numerical se-
ries. This move also involves analysis into creating better functionality that can allow the machine
learning system to develop correct inferences.
• Teach the machine learning algorithm to identify malware by utilising the functionality we’ve
collected.
• Test the solution to such data not used in the practice scenarios to determine how effectively our
recognition method performs.
Let us address all of these measures in further depth in the subsequent pages.
Machine learning indicators succeed or fail based on the testing data given. The capacity of the mali-
cious detector to identify malicious binaries strongly relies on the amount and consistency of the testing
instances you include. Be interested in consuming a lot of time collecting training data while developing
machine learning detection systems, so the more samples you provide to the device, the more effective
it will be. The consistency of the teaching samples is, therefore, quite critical. The vulnerabilities and
benignware that you are accumulating could represent the variety of vulnerabilities and benignware that
you anticipate your scanner to recognise whenever you request to determine if recent documents are
harmful or benign.
For e.g., if you’d like to identify vulnerabilities from either a particular potential risk community,
you need to gather quite so many malicious files as practicable from a certain community for use in
your device training. If you aim to identify a large group of malware (like ransomware), it is crucial to
gather many more qualitative data of such a category as practicable.
Simultaneously, the harmless practice sets that you supply to the device must represent the sort of
benign documents that you would expect your detection system to examine until it is deployed. For
example, when you focus on malware detection on a network connection, you can train your machine
with extensive testing of benignware that graduates and college staff are using to prevent false positives.
52
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
These benign instances could involve video games, text editors, personalised applications created by the
institution IT unit, and other safety programmes.
Extraction Features
To differentiate data as safe or unsafe, we prepare machine learning models by presenting the character-
istics of application binaries; these were all file characteristics that will allow the machine to differentiate
between safe and unsafe data. For e.g., below are a few characteristics that may be used to decide if a
document is safe or unsafe:
• If it is electronically signed
• Existence of deformed headers
• The existence of encrypted information
• If the programme has also been presented on over 100 machines.
To achieve this environment, the files must be taken out. For, e.g., we could build programmes to
decide if a document is electronically signed, has incorrect information, is compressed etc. In defence
data science, we also use many functionalities in machine learning detection systems. For, e.g., we
could build functionality for any API request throughout the Win32 API so that a binary will include
that functionality and possess the specific API call.
Our aim should be to pick the characteristics that produce the best reliable outcomes. This segment
points forth several of the general principles to be observed.
First, when choosing characteristics, pick those that reflect your most robust assumption about what
would allow a machine learning program to differentiate unsafe files against safe ones. For e.g., the
“includes encrypted information” function may be a nice malware flag since we recognise that malware
always includes encrypted information, and we are assuming that benignware would have encrypted
information more seldom. The advantage of machine learning is that when this theory is incorrect and
benignware includes encrypted information as much as malware would, the program can indeed dis-
regard this function. If the idea is correct, the device can know how to utilise the “includes encrypted
information” function to identify malware.
Second, don’t include too many functions further that range of options will get so huge, especially
compared to the amount of practice set for the detection technique. That’s what machine learning re-
searchers term the “Dimensionality Curse.” For e.g., when you have 1,000 characteristics and just 1,000
samples of training, the odds are that you may not have good examples of coaching to tell your machine
learning model how each function means to a provided binary. Stats informs us that this is necessary to
offer the device a few characteristics compared to the number of practice samples you have accessible,
and just let it shape possibly the best assumptions regarding which functionality really imply malware.
Ultimately, make sure that the applications reflect various possibilities on what comprises vulnerability
or benignware. For e.g., you can want to create encrypted communication characteristics, like when a
document uses encrypted communication API calls or a Public Key Infrastructure (PKI), but ensure you
53
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
can use non-encryption-related characteristics to protect your decisions. That means, if the code is unable
to identify malware depending on one form of function, it may still identify it by utilising other functions.
After taking out functions from the testing binaries, it’s essential to teach your machine learning models.
Because this appears like, computationally, it relies mainly on the machine learning method you use.
For e.g., teaching the decision tree method requires a separate learning algorithm than practising the
logistic regression methodology. Luckily, both machine learning detection systems have the same gen-
eral design. You supply them with learning data that includes functionalities from the test binaries and
related identifiers that say the program which repositories are vulnerabilities and what are benignware.
The methods then train to assess whether or not fresh, prior unknown binaries are harmful or benevolent.
The teaching is addressed in more depth further in this chapter
This textbook focuses on the category of machine learning techniques defined as controlled machine
learning techniques. To use machine learning models for malware detection, we prepare models by
telling them which instances are harmful and harmless. Another type of machine learning techniques,
unsupervised methods, doesn’t enable us to recognise which instances are harmful or benevolent in our
practice collection. Malware detection algorithms are far less successful at identifying harmful malware,
and we would not include them in the textbook.
Testing artificial intelligence frameworks and models. After you have learned the machine learning
method, you have to verify how reliable it is. You achieve it by executing a professional data system
that you’ve not learned and seeing how well it decides if or not their samples are harmful or benevolent.
In defence, we usually teach our machines examples that we’ve collected until that moment in time.
Afterwards, we evaluate samples that we’ve encountered after the last moment in time, calculate how
our programs can identify new malware, and assess how well our machines can prevent generating false
positivity on a recent dataset. Many machine learning analysis requires lots of experiments like result
shows: we build a machine learning framework, evaluate it, and then modify it, improve it once again,
and evaluate it once more, perform the process until we are glad by the performance.
Let’s explore how some machine learning algorithms operate. This is the most challenging part of the
chapter but satisfying if you interpret it correctly. Throughout this discussion, we discuss the coherent
principles that encompass these frameworks; then we will move through each method in depth.
Two basic geometric concepts will allow you to recognise all machine learning-based identification
techniques: the concept of a geometrical space function as well as the concept of Judgement limits. You
identify malware binaries on one side of the judgement border and benignware binaries on the other
side. As we use a machine learning method to find data as harmful or harmless, we automatically extract
functions which we can put tests in a function room, and afterwards, we test what side of the judgement
limit the tests are using to decide if the samples are harmful or harmless.
This geometry of knowing the spaces and the limits of options is valid for structures running in one,
two, or 3-dimensional structures (characteristics), but it often encompasses function environments with
billions of dimensions. In contrast, billion-plus dimensional spaces cannot be visualised or conceived.
We’ll stick to two-dimensional illustrations in this section to keep them more accessible to visualise.
54
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
Still, please note that real security machine learning applications almost all use thousands, millions, or
billions of dimensions, as well as the core principles we present in a two-dimensional setting are for
practical cases of more than two dimensions.
Build a toy malware detection issue to explain the concept of a judgement limit in a function space.
Assume we have such a classification model of malicious and harmless tests. Now suppose we derive
the below two attributes within each binary: the amount of the document which tends to be compressed,
as well as the number of questionable operations within each binary source.
CONCLUSION
This chapter provides a comprehensive analysis of the literature relating to machine learning methods
used for malware identification and categorisation. The end of this article concerns a total of 67 aca-
demic papers on malware identification on Windows platforms. These experiments are evaluated using
different critical variables such as the attributes of the databases, classification algorithms, properties of
the studies and the analytical tasks. Our analysis made four significant contributors to this report. The
process of feature extraction, selection and reduction to match model is defined. In this sense, there are
three critical types: (1) static approaches, (2) dynamic approaches, and (3) hybrid approaches.
State-of-the-art information extraction techniques utilise a static-based methodology where no cod-
ing is executed. In addition, dynamic-based approaches provide certain methods that target runtime
features of malware operation. This hybrid method incorporates both static and dynamic elements of
the system. This chapter adds to the research on malware identification by deep learning by presenting a
comprehensive overview of the different methods focused on the network design and how its input data
are structured. Deep neural network methods are classified according to the form of the input.
This chapter discusses approaches that utilise upwards of one modality of the input to identify mal-
ware in an automated manner. Hybrid approaches include both early and late fusion systems. We will
address the idea drift problem and adversarial learning. In addition, we would address the state of the
measurements being used by the research community to test their processes and the question of clas-
sification imbalance. You have already heard about the benefits and disadvantages of dynamic analysis.
Now that you’ve studied how to do simple dynamic research, you’re able to look deeper into the data
science of malware. The majority of this textbook reflects on the performance of malware data science
on dynamic malware data processing.
55
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
REFERENCES
Ab Malek, M. S. B., Ahmadon, M. A. B., Yamaguchi, S., & Gupta, B. B. (2016, October). On privacy
verification in the IoT service based on PN 2. In 2016 IEEE 5th Global Conference on Consumer Elec-
tronics (pp. 1-4). IEEE.
Ahmadi, M., Ulyanov, D., Semenov, S., Trofimov, M., & Giacinto, G. (2016, March). Novel feature
extraction, selection and fusion for effective malware family classification. In Proceedings of the sixth
ACM conference on data and application security and privacy (pp. 183-194). ACM.
Al-Qerem, A., Alauthman, M., Almomani, A., & Gupta, B. B. (2020). IoT transaction processing through
cooperative concurrency control on fog–cloud computing environment. Soft Computing, 24(8), 5695–5711.
Athiwaratkun, B., & Stokes, J. W. (2017, March). Malware classification with LSTM and GRU language
models and a character-level CNN. In 2017 IEEE International Conference on Acoustics, Speech and
Signal Processing (ICASSP) (pp. 2482-2486). IEEE.
Chhabra, M., Gupta, B., & Almomani, A. (2013). A novel solution to handle DDOS attack in MANET.
Journal of Information Security, 4(3), 34631. Advance online publication. doi:10.4236/jis.2013.43019
Dahl, G. E., Stokes, J. W., Deng, L., & Yu, D. (2013, May). Large-scale malware classification using
random projections and neural networks. In 2013 IEEE International Conference on Acoustics, Speech
and Signal Processing (pp. 3422-3426). IEEE.
Ghoneim, A., Muhammad, G., Amin, S. U., & Gupta, B. (2018). Medical image forgery detection for
smart healthcare. IEEE Communications Magazine, 56(4), 33–37.
Gibert, D., Mateu, C., & Planes, J. (2019, July). A hierarchical convolutional neural network for malware
classification. In 2019 International Joint Conference on Neural Networks (IJCNN) (pp. 1-8). IEEE.
Gibert, D., Mateu, C., & Planes, J. (2020). The rise of machine learning for detection and classification
of malware: Research developments, trends and challenges. Journal of Network and Computer Applica-
tions, 153, 102526.
Gibert, D., Mateu, C., Planes, J., & Vicens, R. (2019). Using convolutional neural networks for clas-
sification of malware represented as images. Journal of Computer Virology and Hacking Techniques,
15(1), 15–28.
Gou, Z., & Yamaguchi, S. (2017). Analysis of various security issues and challenges in cloud comput-
ing environment: a survey. In Identity Theft: Breakthroughs in Research and Practice (pp. 221-247).
IGI Global.
Gupta, B. B., Gupta, S., Gangwar, S., Kumar, M., & Meena, P. K. (2015). Cross-site scripting (XSS)
abuse and defense: Exploitation on several testing bed environments and its defense. Journal of Informa-
tion Privacy and Security, 11(2), 118–136.
Gupta, S., & Gupta, B. B. (2015, May). PHP-sensor: a prototype method to discover workflow viola-
tion and XSS vulnerabilities in PHP web applications. In Proceedings of the 12th ACM International
Conference on Computing Frontiers (pp. 1-8). ACM.
56
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
Gupta, S., & Gupta, B. B. (2016). XSS-SAFE: A server-side approach to detect and mitigate cross-site
scripting (XSS) attacks in JavaScript code. Arabian Journal for Science and Engineering, 41(3), 897–920.
Gupta, S., & Gupta, B. B. (2016). JS‐SAN: Defense mechanism for HTML5‐based web applications
against javascript code injection vulnerabilities. Security and Communication Networks, 9(11), 1477–1495.
Gupta, S., & Gupta, B. B. (2017). Detection, avoidance, and attack pattern mechanisms in modern web
application vulnerabilities: Present and future challenges. International Journal of Cloud Applications
and Computing, 7(3), 1–43.
Huang, W., & Stokes, J. W. (2016, July). MtNet: a multi-task neural network for dynamic malware
classification. In International conference on detection of intrusions and malware, and vulnerability
assessment (pp. 399-418). Springer. 10.1007/978-3-319-40667-1_20
Jerbi, W., Guermazi, A., & Trabelsi, H. (2020). A novel secure routing protocol of generation and man-
agement cryptographic keys for wireless sensor networks deployed in internet of things. International
Journal of High Performance Computing and Networking, 16(2-3), 87–94.
Kancherla, K., & Mukkamala, S. (2013, April). Image visualization based malware detection. In 2013
IEEE Symposium on Computational Intelligence in Cyber Security (CICS) (pp. 40-44). IEEE.
Kolosnjaji, B., Zarras, A., Webster, G., & Eckert, C. (2016, December). Deep learning for classification
of malware system call sequences. In Australasian Joint Conference on Artificial Intelligence (pp. 137-
149). Springer. 10.1007/978-3-319-50127-7_11
Krčál, M., Švec, O., Bálek, M., & Jašek, O. (2018). Deep convolutional malware classifiers can learn
from raw executables and labels only. Academic Press.
Letteri, I., Penna, G. D., & Gasperis, G. D. (2019). Security in the internet of things: Botnet detection
in software-defined networks by deep learning techniques. International Journal of High Performance
Computing and Networking, 15(3-4), 170–182.
Mani, N., Moh, M., & Moh, T. S. (2021). Defending deep learning models against adversarial attacks.
International Journal of Software Science and Computational Intelligence, 13(1), 72–89.
Muna, A. H., Moustafa, N., & Sitnikova, E. (2018). Identification of malicious activities in industrial
internet of things based on deep learning models. Journal of Information Security and Applications,
41, 1-11.
Nataraj, L., Karthikeyan, S., Jacob, G., & Manjunath, B. S. (2011, July). Malware images: visualization
and automatic classification. In Proceedings of the 8th international symposium on visualization for
cyber security (pp. 1-7). Academic Press.
Prasse, P., Machlica, L., Pevný, T., Havelka, J., & Scheffer, T. (2017, September). Malware detection by
analysing encrypted network traffic with neural networks. In Joint European Conference on Machine
Learning and Knowledge Discovery in Databases (pp. 73-88). Springer. 10.1007/978-3-319-71246-8_5
Raff, E., Barker, J., Sylvester, J., Brandon, R., Catanzaro, B., & Nicholas, C. (2017). Malware detection
by eating a whole exe. arXiv preprint arXiv:1710.09435.
57
Deep-Learning and Machine-Learning-Based Techniques for Malware Detection and Network Security
Rezende, E., Ruppert, G., Carvalho, T., Ramos, F., & De Geus, P. (2017, December). Malicious software
classification using transfer learning of resnet-50 deep neural network. In 2017 16th IEEE International
Conference on Machine Learning and Applications (ICMLA) (pp. 1011-1014). IEEE.
Ronen, R., Radu, M., Feuerstein, C., Yom-Tov, E., & Ahmadi, M. (2018). Microsoft malware classifica-
tion challenge. arXiv preprint arXiv:1802.10135.
Saxe, J., & Berlin, K. (2015, October). Deep neural network based malware detection using two dimen-
sional binary program features. In 2015 10th International Conference on Malicious and Unwanted
Software (MALWARE) (pp. 11-20). IEEE. 10.1109/MALWARE.2015.7413680
Saxe, J., & Sanders, H. (2018). Malware Data Science: Attack Detection and Attribution. No Starch Press.
Singh, A., & Kumar, R. (2021). A Two-Phase Load Balancing Algorithm for Cloud Environment. In-
ternational Journal of Software Science and Computational Intelligence, 13(1), 38–55.
Solis, D., & Vicens, R. (2017, October). Convolutional neural networks for classification of malware
assembly code. In Recent Advances in Artificial Intelligence Research and Development: Proceedings
of the 20th International Conference of the Catalan Association for Artificial Intelligence, Deltebre,
Terres de L’Ebre, Spain, October 25-27, 2017 (Vol. 300, p. 221). IOS Press.
Tewari, A., & Gupta, B. B. (2017). A lightweight mutual authentication protocol based on elliptic curve
cryptography for IoT devices. International Journal of Advanced Intelligence Paradigms, 9(2-3), 111–121.
Wang, H., Li, Z., Li, Y., Gupta, B. B., & Choi, C. (2020). Visual saliency guided complex image retrieval.
Pattern Recognition Letters, 130, 64–72.
Zhang, X., Zhao, J., & LeCun, Y. (2015). Character-level convolutional networks for text classification.
arXiv preprint arXiv:1509.01626.
Zhang, Z., Sun, R., Zhao, C., Wang, J., Chang, C. K., & Gupta, B. B. (2017). CyVOD: A novel trinity
multimedia social network scheme. Multimedia Tools and Applications, 76(18), 18513–18529.
58
59
Chapter 4
The Era of Advanced
Machine Learning and
Deep Learning Algorithms
for Malware Detection
Kwok Tai Chui
Hong Kong Metropolitan University, Hong Kong
Miltiadis D. Lytras
Deree College — The American College of Greece, Greece
Chien-wen Shen
National Central University, Taiwan
ABSTRACT
Software has been the essential element to computers in today’s digital era. Unfortunately, it has expe-
rienced challenges from various types of malware, which are designed for sabotage, criminal money-
making, and information theft. To protect the gadgets from malware, numerous malware detection
algorithms have been proposed. In the olden days there were shallow learning algorithms, and in recent
years there are deep learning algorithms. With the availability of big data for training of model and
affordable and high-performance computing services, deep learning has demonstrated its superiority
in many smart city applications, in terms of accuracy, error rate, etc. This chapter intends to conduct a
systematic review on the latest development of deep learning algorithms for malware detection. Some
future research directions are suggested for further exploration.
DOI: 10.4018/978-1-7998-7789-9.ch004
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
INTRODUCTION
Computing tools and smartphones have played an impactful role towards smart city vision in recent
decades (Ficco, Esposito, Xiang, & Palmieri, 2017; Rose, Raghuram, Watson, & Wigley, 2021). Ac-
cording to the Statista (Technology Markets: Software, 2021), as shown in Figure 1, there is a steady
growth rate of around 7.1-7.7% in the revenue of software development from 2017 to 2025, except the
historical low 2.6% in 2020 and bounced back to 9% in 2021 during pandemic. The projection could be
altered depending on the deployment of 5G and development of 6G (Stergiou, Psannis, & Gupta, 2020).
Intuitively, the more the number of software linking to gadgets, the more the number of malware at-
tacks. Yet, numerous types of malware have been developed such as scareware, wiper, rogue software,
adware, spyware, ransomware, Trojan horses, worms, and computer viruses (Kumar, 2020; Rendell,
2019). Surprisingly, the global yearly malware attacks (SonicWall, 2021) does not follow an increasing
trend, as shown in Figure 2. From 2015 to 2017, the percentage changes in the number of malware at-
tacks are -3.7% and 8.9%, respectively. There was a notably increment by 22.1% from 2017 to 2018 and
slightly decrement by 5.7% from 2018 to 2019. Compared the last two recorded periods from 2019 to
2020, a significant drop (43.4%) in the number of malware attacks was observed. The key explanation
to the drop of the malware attacks is malware detection algorithms which can detect malware and thus
avoid the damage of gadgets.
A lot of traditional machine learning algorithms was employed for malware detection in literature,
including decision tree, Naïve Bayes, support vector machine, K-nearest neighbour, Bayseian network,
multi-layer perception, J48, and random forest, (Jerlin, & Marimuthu, 2018; Li et al., 2018; Narudin,
Feizollah, Anuar, & Gani, 2016). There is room for improvement in terms of accuracy. Owning to the
fact that a large amount of data is available as training dataset, attention is drawn into deep learning
which can further enhance the accuracy of the detection model.
Figure 1. The worldwide statistics on the revenue of software development (by segment).
60
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Figure 2. The global yearly number of malware attacks between 2015 and 2020.
The research contributions are two-fold (i) a systematic review has been conducted to review the
latest development of deep learning-based malware detection; and (ii) Future research directions on the
enhancement of detection models.
This chapter is structured as follows. Section 2 introduces typical deep learning algorithms. This
is followed by the systematic review in the deep learning algorithms for malware detection. At last, a
conclusion is drawn.
Ten typical deep learning algorithms namely convolutional neural network (CNN), restricted Boltzmann
machine (RBM), deep belief network (DBN), autoencoders (ATE), self organizing map (SOM), mul-
tilayer perceptron (MLP), generative adversarial network (GAN), recurrent neural network (RNN),
gated recurrent unit (GRU), and long short-term memory network (LSTM), are briefly introduced with
background information and their variants proposed in recent years. Table 1 summarizes the number of
publications related to each deep learning algorithms in 2016-2021 (up to 10 August 2021) using the
document search (TITLE-ABS-KEY) in Scopus. The Year 2021 is for reference only because there are
four more months before the end of 2021.
All deep learning algorithms (except RBM) receive increasing attention based on the increasing number
of publications from 2016 to 2020. The average annual growth in the nine algorithms are 222%, 31%,
113%, 4%, 54%, 4560%, 99%, 538%, and 160% for CNN, DBN, ATE, SOM, MLP, GAN, RNN, GRU,
and LSTM, respectively. The remarkable change in the adoption of GAN is because of the effectiveness
61
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
in generating new and synthetic training data for model enhancement (Chui, Liu, Zhao, & De Pablos,
2020; Zhang, Sindagi, & Patel, 2019).
Table 1. Number of publications with Scopus indexing in selected deep learning algorithms from 2016
to 2021 (up to 10 August 2021).
Number of publications
Deep learning
algorithm 2021 (up to 10
2016 2017 2018 2019 2020
Aug.)
CNN 2934 6607 13159 22776 29010 17503
RBM 301 331 324 355 329 171
DBN 357 489 585 726 800 449
ATE 216 378 625 973 1192 683
SOM 872 881 960 925 1012 555
MLP 798 896 1229 1794 2513 1684
GAN 24 291 1360 3158 4402 2640
RNN 1424 2345 3909 5988 7076 3895
GRU 46 126 323 720 1036 681
LSTM 1192 1907 3865 6953 8838 5475
Many review articles have summarized the performance of CNN models for computer vision (Dhillon,
& Verma, 2020), natural language processing (Wu et al., 2020), and remote sensing (Ma et al., 2019),
which revealed the superiority of CNN in these domains. The applications align with the biological
process (Vardhana, Arunkumar, Lasrado, Abdulhay, & Ramirez-Gonzalez, 2018).
Taking classification problem as example, which is comprised of feature learning and model construc-
tion. For feature learning, there are three major components namely convolution filter/kernel, activation
function, and pooling. For readers who are expertise in mathematics or digital signal processing, the
convolution of matrix is equivalent to the concept of cross-correlation. When it comes to the model
construction, flattening, fully-connected layer, and softmax function are major components.
As a form of generative stochastic artificial neural network, RBM restricts the connections between hid-
den units so that the training time can be reduced. The most basic architecture is a two-layered neural
networks with one input layer followed by one hidden layer. In recent years, researchers prefer the selec-
tions of other deep learning algorithms instead of RBM owning to the high performance requirement in
today’s complex research topics (Larochelle, Mandel, Pascanu, & Bengio, 2012).
Some improved variants of RBM have been proposed in recent years. To improve the accuracy and
training time, fuzzy logic was introduced to eliminate the redundancy of data (Lü, Meng, Chen, &
62
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Wang, 2019). Another work suggested the adoption of AdaBoost backward propagation to enhance the
optimization process (Liu, Sun, Kornhauser, Sun, & Sangwa, 2019).
DBN is usually formulated with multiple RBMs. The architecture is structured as directed acyclic graph,
containing a list of stochastic and weighted variables. The training of model can be simplified by intro-
ducing unbiased sampling where there is no interaction between neurons (Movahedi, Coyle, & Sejdić,
2017). However, the type of DBN is defined as generative probabilistic model. With strong relation
to RBMs, DBN is considered as one of the early deep learning-based model. Likewise, DBN receives
relatively less attention compared with other deep learning algorithms.
To improve the generalization, robustness, and accuracy of the model, a compensation of time varying
delay was added to the input layer of DBN (Hao, Wang, Shan, & Zhao, 2019). When it comes to large-
scale datasets, traditional computing framework may not be feasible to manage the training of model in
a reasonable timeframe. A distributed parallel approach was introduced to allocate the computational
load with high-performance clusters (Shi, Zhang, Zhang, & Hu, 2020).
Autoencoders (ATE)
ATE is comprised of three components, encoding algorithm, decoding algorithm, and loss function. It
is usually adopted to reduce the dimensionality of features. Therefore, many previous works incorpo-
rated ATE with deep learning algorithms, for instance, CNN (Toğaçar, Ergen, & Cömert, 2020), MLP
(Ahmadlou et al., 2021), GAN (Wang, Sun, & Jin, 2020), and LSTM (Wang et al., 2021). ATE can be
easily transformed into deep learning by the extension of single layer to multi-layer ATE.
There are other variants of ATE include variational ATE (Zavrak, & Iskefiyeli, 2020), concrete ATE
(Chow et al., 2020), contractive ATE (Diallo et al., 2021), denoising ATE (Ashfahani, Pratama, Lughofer,
& Ong, 2020), and sparse ATE (Mienye, Sun, & Wang, 2020).
SOM reduces the complexity of model by representing higher dimensional data to lower dimensional
data (named map) with a formulation of unsupervised learning. In addition, the topological proper-
ties of the data have been preserved due to neighbourhood function. The idea of SOM ensures various
modules of the network gives comparable reaction to inputs. This can be correlated to eyes, ears, skin,
etc. of human beings. The weights of the nodes are defined by best matching unit (Bernard, Hueber, &
Girau, September 2020).
In general, there are two types of SOMs, growing hierarchical SOM (GH-SOM) and hierarchical SOM
(H-SOM) (Qu et al., 2021). GH-SOM takes the advantages in self-learning, dynamic self-adaptability,
and low computing latency (Vasighi, & Amini, 2017) whereas H-SOM represents hierarchical data ef-
fectively with low computational overhead (Aghajari, & Chandrashekhar, 2017).
63
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
It is often confused by mixing the technique of feedforward artificial neural network with MLP. However,
the multiple layers of perceptron are crucial for MLP where the feature vector is attached to weight vec-
tor. Fundamentally, MLP is comprised of three types of layers namely input, hidden, and output layers.
Updating the weights has become important to control the error of model. Learning rate can limit and
relieve the adjustment of weights between epochs (Takase, Oyama, & Kurihara, 2018). A latest work has
suggested the addition of weight decay or L2-regularization (Richemond, & Guo, 2019).
Although we are living in a big data era where tremendous amount of data is generating by humans,
sensors, devices, or generally speaking internet-of-things network, data collection of some groups of
samples is challenging attributable to limited budget and rare cases in nature. A typical example is
fault samples of machines that suffers from downtime during data collection and requires maintenance
afterwards (Lei et al., 2020).
Some variants of GANs have been proposed to improve the robustness in noise and repentance of
generated data. Examples are conditional GAN, cGAN (Liao, Lin, Zhao, & Gabbouj, 2019), information
maximized generative adversarial networks, infoGAN (Gong, Xu, & Lei, 2020), and auxiliary classifier
GAN, acGAN (Wang et al., 2019).
The inputs and outputs in traditional neural network do not depend on each other and one input maps to
one output. In contrast, the outputs of RNN depend on previous input sequences and there is no restric-
tion on the mapping between inputs and outputs. However, RNN suffers from the challenges of gradient
explosion and vanishing gradient (Yue, Fu, & Liang, 2018).
In general, there are four classes of RNN namely one-to-one (Xiang et al., 2018), one-to-many (Yang,
Liu, Qu, Sang, & Lv, 2021), many-to-one (Dadoun, & Troncy, 2020), and many-to-many (Das, Koperski,
Bremond, & Francesca, November 2018).
GRU is a RNN variant that improves the performance of short-term memory. The regulation of infor-
mation is managed by hidden state. The update and reset gates help to resolve the issue of vanishing
gradient. The former is related to the information from the past moving forward to the future whereas
the latter is related to the information from the past being ignored.
There are two types of architectures namely minimal gated unit (Dong, Du, & Yan, 2019) and fully
gated unit (Khan, Wang, & Ngueilbaye, 2021). Some research studies suggested GRU outperforms RNN
and LSTM in small-scale datasets (Gruber, & Jockisch, 2020).
64
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Both GRU and LSTM can manage the issue of vanishing gradient. The complexity of LSTM is higher
with three gates namely input, forget, and output gates, and thus requiring more computing power. Based
on Table 1, LSTM is more famous compared with RNN and GRU. One of the reasons could be due to
the performance in general applications (Fischer, & Krauss, 2018; Yuan, Li, & Wang, 2019).
We use Scopus to examine the number of publications of the abovementioned 10 deep learning algo-
rithms for malware detection which is summarized in Table 2. The ranking is similar to the results in
Table 1 for general applications.
Table 2. Number of publications of typical deep learning algorithms for malware detection using Scopus.
The articles with top number of citations in each of the deep learning algorithms are selected for in-depth
discussion. Table 3 summarizes the brief information of the shortlisted articles.
In the field of anomaly detection for computer network, traffic classification is important to manage
the network security. In (Wang, Zhu, Zeng, Ye, & Sheng, 2017, January), a CNN-based traffic classifica-
tion model was implemented. The key steps are (i) generating session and flow of the network traffic; (ii)
removing emptied or duplicated files; (iii) sanitizing the trace; (iv) trimming to obtain uniform length;
(v) generating images; (vi) generating IDX files; and (vii) implementing CNN models with IDX files.
Performance evaluation showed an average accuracy of 99.4%.
65
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
A review article (Berman, Buczak, Chavis, & Corbett, 2019) has summarized the results of nine
articles using RBM for malware detection with nine benchmark datasets. Many of the works achieved
accuracy over 95%. The extensive review for other deep learning algorithms such as ATE, CNN, and
RNN are included in the article.
A DBN-based approach was proposed to generate and classify malware signature (David, & Netan-
yahu, 2015, July). The sandbox contains the characteristics of the programs which serve as binary vector.
Six types of malware were selected for performance evaluation of the classifier namely DarkComet,
Andromeda, Cidox, SpyEye, Carberp, and Zeus. The analysis revealed that the malware classification
accuracy was over 98%.
A generic ATE-approach was presented for malware and intrusion detection (Yousefi-Azar, Varad-
harajan, Hamey, & Tupakula, 2017, May). It takes three major advantages (i) unique pre-training; (ii)
minimal dimensions of feature vector; and (iii) unique fine-tuning phase. It was tested with four classi-
fiers, XGBoost, SVM, k-NN, and naïve Bayes, where accuracy was in descending order.
Differed from the work (Wang, Zhu, Zeng, Ye, & Sheng, 2017, January), the SOM-based approach
captured the fuzzy boundary of the machine activity data, including malicious and normal samples (Bur-
nap, French, Turner, & Jones, 2018). Besides, various types of cross-validation were evaluated which
influenced the performance of the classification accuracy. The proposed approach outperformed SVM,
MLP, Bayesian network, and random forest by 8.4-37.7%.
Four algorithms namely SVM, decision tree, naïve Bayes, and k-NN were compared with MLP
for malware detection (Firdausi, Erwin, & Nugroho, 2010, December). Decision tree with J48 setting
achieved the highest accuracy of about 97% as it performed better in feature extraction.
The work (Pascanu, Stokes, Sanossian, Marinescu, & Thomas, 2015, April) combined echo state
network, RNN, and logistic regression for malware classification. The true positive rate improvement
was 98% compared with traditional event model. RNN, GRU, and LSTM were used for malware detec-
tion (Athiwaratkun, & Stokes, 2017, March). Results showed that LSTM achieved the highest accuracy
and improved the true positive rate by 31%. A deep RNN was proposed for malware detection in an
internet of things network and reached an accuracy of 98% (HaddadPajouh, Dehghantanha, Khayami,
& Choo, 2018).
66
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Authors would like to suggest some more articles as future research directions, including cross-site script-
ing detection (Chaudhary, Gupta, Chang, Nedjah, & Chui, 2021), cloud security (Gou, Yamaguchi, &
Gupta, 2017), attack pattern recognition (Gupta, & Gupta, 2017), surveillance system (Chui, Vasant, &
Liu, 2019), distributed denial-of-service attack (Chhabra, Gupta, & Almomani, 2013; Gaurav, Gupta, Hsu,
Yamaguchi, & Chui, 2021, January), fog computing (Hallappanavar, & Birje, 2021), multimedia social
network (Zhang, Sun, Zhao, Wang, Chang, & Gupta, 2017), mutual authentication protocol (Tewari, &
Gupta, 2020), and IT threats (Ordóñez de Pablos, Almunawar, Chui, & Kaliannan, 2021).
Challenges
• Limited evaluation of malware detection models using small-scale datasets. In Table 4, the sample
sizes of the datasets are not large-scale to reflect the practicality in real-world application
• It is desired to implement an algorithm fits the malware detection of all types of malware. This is
to ensure a light hardware requirement to perform the computational tasks
• Malware attack can lead to the lost of huge amounts of money. The ultimate goal of the detection
accuracy is 100%. Ensemble learning and boosting algorithm are potential approaches to further
enhance the performance of the model.
CONCLUSION
In this chapter, we have studied ten typical deep learning algorithms in all disciplines and concluded that
CNN, GAN, RNN, and LSTM are the most widely used algorithms. Attention is drawn into malware
67
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
detection in which articles of each of the deep learning algorithm with highest number of citations are
shortlisted for further discussion. Although various existing works have resulted a high accuracy, true
positive rate, and true negative rate of the malware detection models, there is room for improvement
because the experiments were under some constraints that may not be fully applied to generic and real-
world scenarios. Several future recommendation readings are suggested as future research directions.
REFERENCES
Aghajari, E., & Chandrashekhar, G. D. (2017). Self-organizing map based extended fuzzy C-means
(SEEFC) algorithm for image segmentation. Applied Soft Computing, 54, 347–363. doi:10.1016/j.
asoc.2017.01.003
Ahmadlou, M., Al‐Fugara, A. K., Al‐Shabeeb, A. R., Arora, A., Al‐Adamat, R., Pham, Q. B., Al-Ansari,
N., Linh, N. T. T., & Sajedi, H. (2021). Flood susceptibility mapping and assessment using a novel deep
learning model combining multilayer perceptron and autoencoder neural networks. Journal of Flood
Risk Management, 14(1), e12683. doi:10.1111/jfr3.12683
Ashfahani, A., Pratama, M., Lughofer, E., & Ong, Y. S. (2020). DEVDAN: Deep evolving denoising
autoencoder. Neurocomputing, 390, 297–314. doi:10.1016/j.neucom.2019.07.106
Athiwaratkun, B., & Stokes, J. W. (2017, March). Malware classification with LSTM and GRU language
models and a character-level CNN. In 2017 IEEE International Conference on Acoustics, Speech and
Signal Processing (ICASSP) (pp. 2482-2486). IEEE. 10.1109/ICASSP.2017.7952603
Berman, D. S., Buczak, A. L., Chavis, J. S., & Corbett, C. L. (2019). A survey of deep learning methods
for cyber security. Information (Basel), 10(4), 122. doi:10.3390/info10040122
Bernard, Y., Hueber, N., & Girau, B. (2020, September). A fast algorithm to find Best Matching Units in
Self-Organizing Maps. In International Conference on Artificial Neural Networks (pp. 825-837). Springer.
Burnap, P., French, R., Turner, F., & Jones, K. (2018). Malware classification using self organising feature
maps and machine activity data. Computers & Security, 73, 399–410. doi:10.1016/j.cose.2017.11.016
Chaudhary, P., Gupta, B. B., Chang, X., Nedjah, N., & Chui, K. T. (2021). Enhancing big data security
through integrating XSS scanner into fog nodes for SMEs gain. Technological Forecasting and Social
Change, 168, 120754. doi:10.1016/j.techfore.2021.120754
Chhabra, M., Gupta, B., & Almomani, A. (2013). A novel solution to handle DDOS attack in MANET.
Journal of Information Security, 4(3), 34631. doi:10.4236/jis.2013.43019
Chow, J. K., Su, Z., Wu, J., Tan, P. S., Mao, X., & Wang, Y. H. (2020). Anomaly detection of defects on
concrete structures with the convolutional autoencoder. Advanced Engineering Informatics, 45, 101105.
doi:10.1016/j.aei.2020.101105
Chui, K. T., Liu, R. W., Zhao, M., & De Pablos, P. O. (2020). Predicting students’ performance with
school and family tutoring using generative adversarial network-based deep support vector machine. IEEE
Access: Practical Innovations, Open Solutions, 8, 86745–86752. doi:10.1109/ACCESS.2020.2992869
68
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Chui, K. T., Vasant, P., & Liu, R. W. (2019). Smart city is a safe city: information and communication
technology–enhanced urban space monitoring and surveillance systems: the promise and limitations. In
Smart Cities: Issues and Challenges (pp. 111-124). Elsevier. doi:10.1016/B978-0-12-816639-0.00007-7
Dadoun, A., & Troncy, R. (2020). Many-to-one Recurrent Neural Network for Session-based Recom-
mendation. arXiv preprint arXiv:2008.11136.
Das, S., Koperski, M., Bremond, F., & Francesca, G. (2018, November). Deep-temporal lstm for daily
living action recognition. In 2018 15th IEEE International Conference on Advanced Video and Signal
Based Surveillance (AVSS) (pp. 1-6). IEEE. 10.1109/AVSS.2018.8639122
David, O. E., & Netanyahu, N. S. (2015, July). Deepsign: Deep learning for automatic malware signature
generation and classification. In 2015 International Joint Conference on Neural Networks (IJCNN) (pp.
1-8). IEEE. 10.1109/IJCNN.2015.7280815
Dhillon, A., & Verma, G. K. (2020). Convolutional neural network: A review of models, methodologies
and applications to object detection. Progress in Artificial Intelligence, 9(2), 85–112. doi:10.100713748-
019-00203-0
Diallo, B., Hu, J., Li, T., Khan, G. A., Liang, X., & Zhao, Y. (2021). Deep embedding clustering based
on contractive autoencoder. Neurocomputing, 433, 96–107. doi:10.1016/j.neucom.2020.12.094
Dong, A., Du, Z., & Yan, Z. (2019). Round trip time prediction using recurrent neural networks with
minimal gated unit. IEEE Communications Letters, 23(4), 584–587. doi:10.1109/LCOMM.2019.2899603
Ficco, M., Esposito, C., Xiang, Y., & Palmieri, F. (2017). Pseudo-dynamic testing of realistic edge-fog
cloud ecosystems. IEEE Communications Magazine, 55(11), 98–104. doi:10.1109/MCOM.2017.1700328
Firdausi, I., Erwin, A., & Nugroho, A. S. (2010, December). Analysis of machine learning techniques
used in behavior-based malware detection. In 2010 second international conference on advances in
computing, control, and telecommunication technologies (pp. 201-203). IEEE. 10.1109/ACT.2010.33
Fischer, T., & Krauss, C. (2018). Deep learning with long short-term memory networks for financial market
predictions. European Journal of Operational Research, 270(2), 654–669. doi:10.1016/j.ejor.2017.11.054
Gaurav, A., Gupta, B. B., Hsu, C. H., Yamaguchi, S., & Chui, K. T. (2021, January). Fog Layer-based
DDoS attack Detection Approach for Internet-of-Things (IoTs) devices. In 2021 IEEE International
Conference on Consumer Electronics (ICCE) (pp. 1-5). IEEE.
Gong, J., Xu, X., & Lei, Y. (2020). Unsupervised specific emitter identification method using radio-
frequency fingerprint embedded InfoGAN. IEEE Transactions on Information Forensics and Security,
15, 2898–2913. doi:10.1109/TIFS.2020.2978620
Gou, Z., Yamaguchi, S., & Gupta, B. B. (2017). Analysis of various security issues and challenges in
cloud computing environment: a survey. In Identity Theft: Breakthroughs in Research and Practice (pp.
221-247). IGI Global. doi:10.4018/978-1-5225-0808-3.ch011
Gruber, N., & Jockisch, A. (2020). Are GRU cells more specific and LSTM cells more sensitive in
motive classification of text? Frontiers in Artificial Intelligence, 3, 40. doi:10.3389/frai.2020.00040
PMID:33733157
69
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Gupta, S., & Gupta, B. B. (2017). Detection, avoidance, and attack pattern mechanisms in modern web
application vulnerabilities: Present and future challenges. International Journal of Cloud Applications
and Computing, 7(3), 1–43. doi:10.4018/IJCAC.2017070101
HaddadPajouh, H., Dehghantanha, A., Khayami, R., & Choo, K.-K. R. (2018). A deep recurrent neural
network based approach for internet of things malware threat hunting. Future Generation Computer
Systems, 85, 88–96. doi:10.1016/j.future.2018.03.007
Hallappanavar, V. L., & Birje, M. N. (2021). A reliable trust computing mechanism in fog computing.
International Journal of Cloud Applications and Computing, 11(1), 1–20. doi:10.4018/IJCAC.2021010101
Hao, X., Wang, Z., Shan, Z., & Zhao, Y. (2019). Prediction of electricity consumption in cement produc-
tion: A time-varying delay deep belief network prediction method. Neural Computing & Applications,
31(11), 7165–7179. doi:10.100700521-018-3540-z
Jerlin, M. A., & Marimuthu, K. (2018). A new malware detection system using machine learning tech-
niques for API call sequences. Journal of Applied Security Research, 13(1), 45–62. doi:10.1080/1936
1610.2018.1387734
Khan, M., Wang, H., & Ngueilbaye, A. (2021). Attention-Based Deep Gated Fully Convolutional End-
to-End Architectures for Time Series Classification. Neural Processing Letters, 1–34.
Kumar, S. (2020). An emerging threat Fileless malware: A survey and research challenges. Cybersecurity,
3(1), 1–12. doi:10.118642400-019-0043-x
Larochelle, H., Mandel, M., Pascanu, R., & Bengio, Y. (2012). Learning algorithms for the classification
restricted boltzmann machine. Journal of Machine Learning Research, 13(1), 643–669.
Lei, Y., Yang, B., Jiang, X., Jia, F., Li, N., & Nandi, A. K. (2020). Applications of machine learning
to machine fault diagnosis: A review and roadmap. Mechanical Systems and Signal Processing, 138,
106587. doi:10.1016/j.ymssp.2019.106587
Li, J., Sun, L., Yan, Q., Li, Z., Srisa-An, W., & Ye, H. (2018). Significant permission identification for
machine-learning-based android malware detection. IEEE Transactions on Industrial Informatics, 14(7),
3216–3225. doi:10.1109/TII.2017.2789219
Liao, K., Lin, C., Zhao, Y., & Gabbouj, M. (2019). DR-GAN: Automatic radial distortion rectification
using conditional GAN in real-time. IEEE Transactions on Circuits and Systems for Video Technology,
30(3), 725–733. doi:10.1109/TCSVT.2019.2897984
Liu, Q., Sun, L., Kornhauser, A., Sun, J., & Sangwa, N. (2019). Road roughness acquisition and clas-
sification using improved restricted Boltzmann machine deep learning algorithm. Sensor Review, 39(6),
733–742. doi:10.1108/SR-05-2018-0132
Lü, X., Meng, L., Chen, C., & Wang, P. (2019). Fuzzy removing redundancy restricted boltzmann
machine: Improving learning speed and classification accuracy. IEEE Transactions on Fuzzy Systems,
28(10), 2495–2509. doi:10.1109/TFUZZ.2019.2940415
70
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Ma, L., Liu, Y., Zhang, X., Ye, Y., Yin, G., & Johnson, B. A. (2019). Deep learning in remote sensing
applications: A meta-analysis and review. ISPRS Journal of Photogrammetry and Remote Sensing, 152,
166–177. doi:10.1016/j.isprsjprs.2019.04.015
Mienye, I. D., Sun, Y., & Wang, Z. (2020). Improved sparse autoencoder based artificial neural network
approach for prediction of heart disease. Informatics in Medicine Unlocked, 18, 100307. doi:10.1016/j.
imu.2020.100307
Movahedi, F., Coyle, J. L., & Sejdić, E. (2017). Deep belief networks for electroencephalography: A
review of recent contributions and future outlooks. IEEE Journal of Biomedical and Health Informatics,
22(3), 642–652. doi:10.1109/JBHI.2017.2727218 PMID:28715343
Narudin, F. A., Feizollah, A., Anuar, N. B., & Gani, A. (2016). Evaluation of machine learning clas-
sifiers for mobile malware detection. Soft Computing, 20(1), 343–357. doi:10.100700500-014-1511-6
Ordóñez de Pablos, P., Almunawar, M. N., Chui, K. T., & Kaliannan, M. (Eds.). (2021). Handbook of
Research on Analyzing IT Opportunities for Inclusive Digital Learning. IGI Global. doi:10.4018/978-
1-7998-7184-2
Pascanu, R., Stokes, J. W., Sanossian, H., Marinescu, M., & Thomas, A. (2015, April). Malware clas-
sification with recurrent networks. In 2015 IEEE International Conference on Acoustics, Speech and
Signal Processing (ICASSP) (pp. 1916-1920). IEEE. 10.1109/ICASSP.2015.7178304
Qu, X., Yang, L., Guo, K., Ma, L., Sun, M., Ke, M., & Li, M. (2021). A survey on the development of
self-organizing maps for unsupervised intrusion detection. Mobile Networks and Applications, 26(2),
808–829. doi:10.100711036-019-01353-0
Rendell, D. (2019). Understanding the evolution of malware. Computer Fraud & Security, 2019(1),
17–19. doi:10.1016/S1361-3723(19)30010-7
Richemond, P. H., & Guo, Y. (2019). Combining learning rate decay and weight decay with complexity
gradient descent-Part I. arXiv preprint arXiv:1902.02881.
Rose, G., Raghuram, P., Watson, S., & Wigley, E. (2021). Platform urbanism, smartphone applications
and valuing data in a smart city. Transactions of the Institute of British Geographers, 46(1), 59–72.
doi:10.1111/tran.12400
Shi, G., Zhang, J., Zhang, C., & Hu, J. (2020). A distributed parallel training method of deep belief
networks. Soft Computing, 24(17), 1–12. doi:10.100700500-020-04754-6
SonicWall. (2021). Annual number of malware attacks worldwide from 2015 to 2020 (in billions). In
Statista. Retrieved 9 August 2021, from https://www.statista.com/statistics/873097/malware-attacks-
per-year-worldwide/
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based big data secure management in the
fog over a 6G wireless network. IEEE Internet of Things Journal, 8(7), 5164–5171. doi:10.1109/
JIOT.2020.3033131
Takase, T., Oyama, S., & Kurihara, M. (2018). Effective neural network training with adaptive learning rate
based on training loss. Neural Networks, 101, 68–78. doi:10.1016/j.neunet.2018.01.016 PMID:29494873
71
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Technology markets: Software, Revenue by segment. (2021). In Statista. Retrieved 9 August 2021, from
https://www.statista.com/statistics/873097/malware-attacks-per-year-worldwide/
Tewari, A., & Gupta, B. B. (2020). Secure Timestamp-Based Mutual Authentication Protocol for IoT
Devices Using RFID Tags. International Journal on Semantic Web and Information Systems, 16(3),
20–34. doi:10.4018/IJSWIS.2020070102
Vardhana, M., Arunkumar, N., Lasrado, S., Abdulhay, E., & Ramirez-Gonzalez, G. (2018). Convolu-
tional neural network for bio-medical image segmentation with hardware acceleration. Cognitive Systems
Research, 50, 10–14. doi:10.1016/j.cogsys.2018.03.005
Vasighi, M., & Amini, H. (2017). A directed batch growing approach to enhance the topology preser-
vation of self-organizing map. Applied Soft Computing, 55, 424–435. doi:10.1016/j.asoc.2017.02.015
Wang, D., Vinson, R., Holmes, M., Seibel, G., Bechar, A., Nof, S., & Tao, Y. (2019). Early detection
of tomato spotted wilt virus by hyperspectral imaging and outlier removal auxiliary classifier genera-
tive adversarial nets (OR-AC-GAN). Scientific Reports, 9(1), 1–14. doi:10.103841598-019-40066-y
PMID:30867450
Wang, H., Peng, M. J., Miao, Z., Liu, Y. K., Ayodeji, A., & Hao, C. (2021). Remaining useful life pre-
diction techniques for electric valves based on convolution auto encoder and long short term memory.
ISA Transactions, 108, 333–342. doi:10.1016/j.isatra.2020.08.031 PMID:32891421
Wang, W., Zhu, M., Zeng, X., Ye, X., & Sheng, Y. (2017, January). Malware traffic classification using
convolutional neural network for representation learning. In 2017 International Conference on Informa-
tion Networking (ICOIN) (pp. 712-717). IEEE. 10.1109/ICOIN.2017.7899588
Wang, Y. R., Sun, G. D., & Jin, Q. (2020). Imbalanced sample fault diagnosis of rotating machinery
using conditional variational auto-encoder generative adversarial network. Applied Soft Computing, 92,
106333. doi:10.1016/j.asoc.2020.106333
Wu, S., Roberts, K., Datta, S., Du, J., Ji, Z., Si, Y., Soni, S., Wang, Q., Wei, Q., Xiang, Y., Zhao, B.,
& Xu, H. (2020). Deep learning in clinical natural language processing: A methodical review. Journal
of the American Medical Informatics Association: JAMIA, 27(3), 457–470. doi:10.1093/jamia/ocz200
PMID:31794016
Xiang, W., Zhang, H., Cui, R., Chu, X., Li, K., & Zhou, W. (2018). Pavo: A rnn-based learned inverted
index, supervised or unsupervised? IEEE Access: Practical Innovations, Open Solutions, 7, 293–303.
doi:10.1109/ACCESS.2018.2885350
Yang, K., Liu, D., Qu, Q., Sang, Y., & Lv, J. (2021). An automatic evaluation metric for Ancient-Modern
Chinese translation. Neural Computing & Applications, 33(8), 3855–3867. doi:10.100700521-020-05216-8
Yousefi-Azar, M., Varadharajan, V., Hamey, L., & Tupakula, U. (2017, May). Autoencoder-based fea-
ture learning for cyber security applications. In 2017 International joint conference on neural networks
(IJCNN) (pp. 3854-3861). IEEE. 10.1109/IJCNN.2017.7966342
72
The Era of Advanced Machine Learning and Deep Learning Algorithms for Malware Detection
Yuan, X., Li, L., & Wang, Y. (2019). Nonlinear dynamic soft sensor modeling with supervised long short-
term memory network. IEEE Transactions on Industrial Informatics, 16(5), 3168–3176. doi:10.1109/
TII.2019.2902129
Yue, B., Fu, J., & Liang, J. (2018). Residual recurrent neural networks for learning sequential representa-
tions. Information (Basel), 9(3), 56. doi:10.3390/info9030056
Zavrak, S., & Iskefiyeli, M. (2020). Anomaly-based intrusion detection from network flow features us-
ing variational autoencoder. IEEE Access: Practical Innovations, Open Solutions, 8, 108346–108358.
doi:10.1109/ACCESS.2020.3001350
Zhang, H., Sindagi, V., & Patel, V. M. (2019). Image de-raining using a conditional generative adver-
sarial network. IEEE Transactions on Circuits and Systems for Video Technology, 30(11), 3943–3956.
doi:10.1109/TCSVT.2019.2920407
Zhang, Z., Sun, R., Zhao, C., Wang, J., Chang, C. K., & Gupta, B. B. (2017). CyVOD: A novel trin-
ity multimedia social network scheme. Multimedia Tools and Applications, 76(18), 18513–18529.
doi:10.100711042-016-4162-z
73
74
Chapter 5
Malware Detection in
Industrial Scenarios Using
Machine Learning and Deep
Learning Techniques
Ángel Luis Perales Gómez
University of Murcia, Spain
ABSTRACT
In the last decades, factories have suffered a significant change in automation, evolving from isolated
towards interconnected systems. However, the adoption of open standards and the opening to the in-
ternet have caused an increment in the number of attacks. In addition, traditional intrusion detection
systems relying on a signature database, where malware patterns are stored, are failing due to the high
specialization of industrial cyberattacks. For this reason, the research community is moving towards
the anomaly detection paradigm. This paradigm is showing great results when it is implemented us-
ing machine learning and deep learning techniques. This chapter surveys several incidents caused by
cyberattacks targeting industrial scenarios. Next, to understand the current status of anomaly detection
solutions, it analyses the current industrial datasets and anomaly detection systems in the industrial
field. In addition, the chapter shows an example of malware attacking a manufacturing plant, resulting
in a safety threat. Finally, cybersecurity and safety solutions are reviewed.
DOI: 10.4018/978-1-7998-7789-9.ch005
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
INTRODUCTION
Nowadays, industry plays a fundamental role in our society since an essential part of the economy is
based on this sector. Therefore, any advance that involves a significant increase in the industrial produc-
tion of factories is associated with an improvement in the economy and, consequently, with the growth
of countries. In this context, industrial processes automation has been the way followed by factories to
increase production without increasing cost.
Decades ago, automation in factories consisted of small isolated elements capable of making
measurements and, based on these measurements, performing certain types of actions. However, new
technologies are being introduced progressively in the industrial ecosystem, facilitating the automation
of processes. In recent years, new terms, such as Industry 4.0 (Lasi et al., 2014), Industrial Internet of
Things (IIoT) (Boyes et al., 2018), and recently, Industry 5.0 and Society 5.0 (Perakovic et al., 2020),
have emerged strongly. In general, these terms are related to each other and refer to introducing new
smart devices in industrial factories. These devices use typical technologies of communication networks,
such as Ethernet or WiFi, to exchange information between them. In addition, more and more factories
are being connected to the Internet (Mirian et al., 2016) to provide new functions such as remote control
or information sharing between factories in different geographical areas. In addition, and based on the
Industry 5.0 paradigm, Artificial Intelligence (AI) techniques are being introduced in industrial scenarios
(Skovelev et al., 2017).
Although Industry 4.0/5.0 comprises many different elements and the identification of its parameters
is crucial (Perakovic et al., 2020), the core part of the factory automation are the Industrial Control
Systems (ICS) that encompass a large number of heterogeneous devices whose goal is to control and
supervise industrial processes. To achieve this goal, ICS comprise devices that operate in both the logical
and physical layers of industrial processes. This is the reason why ICS are also known as Cyber-Physical
Systems (CPS). Devices in the logical layer govern the system behavior, while devices in the physical
layer, such as controllers and sensors, interact with the physical world.
ICS control and supervise the industrial processes from manufacturing industries to critical infra-
structures such as power-grids or gas pipelines. They are placed from the second level to below in the
automation pyramid, as shown in Figure 1. In the zero level of the pyramid, we found low-level devices
nearest to the controlled process; in other words, the physical layer. Among these devices, actuators
and sensors are responsible for getting information from the environment and making actions over the
physical world. In the upper level, the Programmable Logic Controllers (PLC) are located. PLC serve
as intermediaries between the lower level (sensors and actuators) and the next upper level. To be spe-
cific, PLC receive commands from the upper level that are transmitted to actuators/sensors. Sensors
and actuators execute the commands received and send back the result to the PLC, which transmits it
to the upper level. In the second level of the automation pyramid, we find the Supervisory Control And
Data Acquisition (SCADA) systems (Boyer, 2019). These systems are responsible for monitoring and
controlling all the devices in the ICS. At this level, we see devices and software closer to traditional
networks than control networks are. For example, devices that incorporate web, database, and histo-
rian servers are located at this level. SCADA collect the information from different PLC and store it in
databases. Most SCADA can manage alarms and display trend graphics and other useful information
shown in Human-Machine Interface (HMI) devices. Operators use these devices to monitor the correct
operation of the processes. If a process malfunction is detected, operators can interact with the process
through the HMI. The last two levels of the automation pyramid are more related to business tasks than
75
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
factory processes. Specifically, the third level is in charge of tracking and gathering accurate, real-time
data about the complete production lifecycle through the Management Execution System (MES). This
level groups all tasks between order release and completion of the product for distribution. Finally, the
fourth level is in charge of production planning and order management through Enterprise Resource
Planning (ERP), a suite of different software that monitors all business levels, including manufacturing,
orders, sales, finances, payroll, among others. Figure 2 illustrates a general ICS architecture and how
the monitoring level is connected with the lower levels.
Since ICS supervise and control the factory production, they need to be available as long as possible.
Otherwise, factory production may be affected, producing economic losses. This need forces many ICS
around the world to remain outdated because updating an ICS component involves halting the process
and replacing the component, thus stopping production. Moreover, this situation causes a large number
of ICS to be made up of obsolete devices that potentially have critical vulnerabilities. In the past, since
factories were isolated environments, this circumstance was not a critical issue. However, in recent
years, the adoption of open standards and the opening to the Internet have exposed these vulnerabilities,
causing an increment in the number of attacks affecting ICS (Miller et al., 2012; Nicholson et al., 2012;
Repository of Industrial Security Incidents).
Most of the cyberattacks are carried out by means of malware that is introduced in the control network.
The most notorious incident was carried out in 2010 by Stuxnet malware, which goal was to sabotage
uranium enrichment plants of Iran. Another important incident dates from 2017 when Wannacry spread
around the world. Wannacry is a type of malware called ransomware that encrypts files in the computer
76
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
hard drive and asks for a ransom in bitcoin. Although Wannacry was not aimed to attack ICS, it reached
many critical infrastructure systems such as hospitals and manufacturer factories.
Due to the attention that industrial scenarios are receiving by attackers, there is also a growing interest
in introducing cybersecurity mechanisms to protect them. One of the most powerful mechanisms to secure
them is called Intrusion Detection Systems (IDS), which are in charge of analyzing the network (NIDS)
or the host (HIDS) to detect cyberattacks. If the IDS detects any unusual activity, it triggers an alarm
to inform the administrators, who will take the appropriate actions to respond to the detected activity.
Due to the specialized attacks that suffer control networks, traditional IDS fail to detect attacks in
the control network context. These IDS are signature-based; in other words, they compare the pattern
77
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
obtained with those stored in their databases (Wu et al., 2008). If the pattern match, it is considered
an attack. For this reason, the research community is now moving towards IDS based on the anomaly
detection (AD) paradigm (Liao et al., 2013) instead of signature-based IDS. AD systems try to find the
boundary between normal and abnormal behaviors.
An anomaly is a pattern that does not match the expected behavior (Chandola et al., 2009). For ex-
ample, an anomaly in an industrial context could be a higher than expected temperature received by a
sensor. In general, an anomaly is hard to define because it can change over time. In a network without
Modbus traffic, which is an industrial protocol for data exchange between control devices, a Modbus
packet can be considered as an anomaly. However, the control network may be expanded in the future
with new devices that exchange data with Modbus protocol. In this case, a Modbus packet would no
longer be considered anomalous, and a redefinition of the anomaly concept would be required.
Many techniques have been developed to be used in an AD context, but recent approaches based on
Machine Learning (ML) (Xin et al., 2018) and Deep Learning (DL) (Vinayakumar et al., 2019) are gain-
ing prominence. Both ML and DL algorithms need data as input to build a model in a first phase called
training, where models try to learn useful structural information underlying the data. When the model
is trained, it must be validated against unseen data to determine its performance. From here, an iterative
process begins in which the hyper-parameters of the model must be tweak until an adequate performance
is reached. Finally, the model trained can be used to predict anomalies from new unseen data.
We can differentiate three approaches regarding ML and DL model training.
• Supervised approach. The model is feed with a labeled dataset. Inside this category, we find a
variety of algorithms, such as linear regression, Support Vector Machine (SVM), Random Forest
(RF), and several Neural Networks (NN) architectures (Hastie et al., 2005), just to name a few. In
this approach, the error between the predicted label and the ground-truth label is used to adjust the
parameters of the model.
• Semi-supervised approach. In this case, it is not necessary to provide a completely labeled da-
taset. It is sufficient to give a dataset containing only normal samples. Local Outlier Factor (LOF)
and One-Class SVM (OCSVM) (Hastie et al., 2005) are just a few examples of semi-supervised
algorithms. These models try to find a border enclosing the normal samples.
• Unsupervised approach. The dataset used in this approach is not labeled at all. Most of the algo-
rithms in this category are based on clustering, like the K-means algorithm. These kinds of tech-
niques try to group samples close to each other to form a cluster. A complex task in this approach
is to find the correct number of clusters, which is another hyper-parameter.
To train, validate, and compare different models, there are datasets available in public repositories.
Even though many of them are available in the context of corporate networks (despite its drawbacks
(Tavallaee et al., 2009), in control networks, it is less frequent to find datasets ready to be used in model
training and validation. The lack of these datasets is motivated by the fact that control network traffic
manages confidential data about the processes carried out in the factory (e.g., the temperature that a
tank must maintain or the speed at which an engine must turn). This secrecy in the industry hinders the
capture of network traffic and, therefore, restricts the availability of industrial datasets.
Although cybersecurity in industrial scenarios has been improved in recent years, we have identified
a series of open challenges that affect to Industry 4.0/5.0 scenarios that need to be solved to secure the
industrial devices on which automation is based. The first open challenge is related to the lack of indus-
78
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
trial-oriented datasets that hinder the training of ML/DL models to be used in AD-based IDS. Moreover,
many of the available industrial datasets are collected from a synthetic scenario instead of a real one,
resulting in not representative datasets. Likewise, the high specialization of ICS makes impossible the
dataset exchange between different scenarios. As a result, a dataset obtained from a water management
ICS is not appropriate for smart grids. The second open challenge is the lack of a standard methodology
to develop AD models to detect cyberattacks specialized in ICS scenarios. The existing solutions apply
different steps without standard criteria. The lack of this common methodology may even render the
researchers’ results invalid because wrong steps are performed. Finally, the third challenge is related to
the lack of systems capable of managing cybersecurity and safety in a unified way to protect both assets
and workers. These two concepts are closely related to each other, and cybersecurity threats can poten-
tially cause safety threats, as we see in a later section. Until now, the management of cybersecurity and
safety has been carried out in a separate way, causing an increase in both the cost and the complexity of
having two different systems. In addition, current safety management systems cannot correctly evaluate
threats caused by a cyberattack, making detection and mitigation difficult.
This chapter presents a review of works in different fields needed to implement automation in a secure
way. In particular, a review of industrial-oriented datasets to train AD-based IDS in industrial scenarios
is performed. Besides, different works and methodologies that used those datasets to train AD-based
IDS in an industrial context are presented. Finally, we review the different solutions in order to manage
cybersecurity and safety in an industrial context.
The chapter is organized as follows. In Section 2, we survey different incidents in ICS produced by
cyberattacks over the years. In Section 3, we discuss the different industrial-oriented datasets avail-
able in order to train AD-based IDS. In Section 4, we present different solutions that are trained using
industrial datasets to detect cyberattacks in industrial scenarios. In Section 5, we present an example
where a botnet controls an industrial device and how it can cause a safety threat. In Section 6, we discuss
solutions that consider both cybersecurity and safety and, specifically, we review SafeMan, which is a
framework that manages cybersecurity and safety in a unified way. Finally, in Section 7, we present the
conclusion of this chapter.
The actors involved in attacks targeting ICS can be of different kinds, from national governments to
disgruntled employees. Their motivations are also wide and different. For example, national governments
might be interested in spying on another nation, while disgruntled employees only look for revenge due
to a dismissal or a reprimand suffered previously. Table 1 summarizes the different actors involved in
attacks on industrial systems.
In the following paragraphs, we discuss several cyberattacks targeting ICS over the years (Miller et
al., 2012; Nicholson et al., 2012; Repository of Industrial Security Incidents). Some of them are collateral
damages caused by worms that hit the Internet and penetrated into the ICS. Others are attacks focused
on the corporate network that directly impact ICS. However, the most damaging attacks are targeted
against devices that control and supervise the factory processes.
Siberian Gas Pipeline Incident. In 1982, the Siberian gas pipeline suffered an explosion caused by
pipeline control software. The reason was that trojan malware was included with the software. The blast
was caused during a pressure test on the pipeline due to an increase in the normal pressure. Russian
79
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
newspaper recognized the explosion, but they denied that it was caused by software. Mainly, the attack
had two effects. On the one hand, it affected the USSR economy during a crucial event, the cold war,
interrupting the gas supply in the USSR. On the other hand, it had a psychological impact giving an
advantage to the USA in the cold war.
Adversary Description
State-run, well organized, and financed. Use foreign service agents to gather classified or critical
Nation States information from countries viewed as hostile or as having an economic, military, or political
advantage.
A group of individuals who attack networks and systems seeking to exploit the vulnerabilities in
Hackers
operating systems or other flaws.
Individuals or groups operating domestically or internationally who represent various terrorist or
Terrorists/Cyberterrorists extremist groups that use violence or the threat of violence to incite fear to coerce or intimidate
governments or societies into succumbing to their demands.
Coordinated criminal activities, including gambling, racketeering, narcotics trafficking, and many
Organized Crime
others. An organized and well-financed criminal organization.
Another facet of the criminal community, which is usually not well organized or financed. Consists
Other Criminal Elements
typically of few individuals or of one individual acting alone.
Foreign and domestic corporations operating in a competitive market and often engaged in the illegal
Industrial Competitors
gathering of information from competitors or foreign governments in the form of corporate espionage.
Angry, dissatisfied individuals with the potential to inflict harm on the smart grid network or related
Disgruntled Employees systems. This can represent an insider threat depending on the current state of the individual’s
employment and access to the systems.
Careless or Poorly Trained Those users who, either through lack of training, lack of concern, or lack of attentiveness, pose a
Employees threat to smart grid systems. This is another example of an insider threat or adversary.
Nuclear Power Plant Incident. In 1992, a computer programmer in the Ignalia Power Reactor Station
(Lithuania) introduced a computer virus in one station to sabotage the reactor, causing the station to shut
down. It is speculated that the virus affects the cooling system causing it to break.
Salt River Project Incident. In 1994, between July 8th and August 31st, Lane Jarret Davis accessed
a computer belonging to the Salt River Project through a dial-up modem. Lane was able to access and
delete data from systems responsible for the monitoring and delivery of water and power customers as
well as personnel and financial data. Salt River Project estimated the loss in $40,000 without regarding
the loss due to productivity.
Omega Engineering Incident. On July 31st, 1994, a worker from the manufacturing plant of Omega
Engineering in Bridgeport, N.J., found malware on his computer after login. The malware removed
all the programs that ran the manufacturing operations of the company. After the incident, the CFO of
Omega declared that the malware removed the software that controlled the production of 25,000 differ-
ent products and customized those products with 500,000 different designs. Omega loss is quantified
as 12 million in damages that breaks down into 10 million in sales and other costs, and 2 million to
reprogram the system, apart from losing its competitive position in the market for high-tech instruments
and measurements.
80
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
SQL Slammer (Moore et al., 2003). In 2003, a new cyber threat was discovered by Michael Bacarella.
SQL Slammer was a tiny worm (just 376 bytes) that exploited a SQL Server buffer overflow vulner-
ability. The worm used the UDP protocol and the port associated with Microsoft SQL Server (1434).
This worm not only affected IT devices and data networks but also control devices and control networks.
The worm replicated and sent itself to a large number of devices. This caused Denial-of-Service (DoS)
attacks in a significant number of networks around the world.
SQL Slammer entered the automation network segment in a drilling company and caused a DoS
attack. The attack caused intermittent network traffic between operator terminals and SCADA servers.
Unfortunately, the attack also affected a PLC and Distributed Control System (DCS) connected to the
SCADA.
In the power industry, there have been several incidents where SQL Slammer blocked the communi-
cations between SCADA and PLC, but the most important one was on January 25th, in the Davis-Besse
nuclear power plant. The infection caused data overload in the network, causing the communication
between computers were blocked.
At first, the plant was safe from the SQL Slammer worm because it was connected to the Internet via
a firewall. The firewall was setup to blocking UDP connection on port 1434 (Microsoft SQL Server).
However, the entry point of the SQL Slammer worm was a T1 line behind the firewall that provided a
path to enter the system. This path bypasses all access control securities applied by the firewall. Once
the worm entered to the internal network, it randomly spread to other hosts whose port 1434 was open.
Another impact of SQL Slammer was in an automotive manufacturer. On June 25th, the worm affected
17 plants, including around 1000 computers. The estimated cost to the company was 150 million dollars.
CSX Train Signalling System Incident. In 2003, another worm spread over the Internet; this time, it
was Sobig. It entered into the CSX network affecting the train signaling system. Around 1:15 a.m., Sobig
shut down signaling, dispatching, and other systems. As a result, passenger and freight train traffic was
stopped as soon as possible, and some trains suffered delays between 15 and 30 minutes.
Sasser (Li et al., 2008). In 2004 a worm named Sasser spread over the Internet. The worm exploited
a buffer overflow vulnerability in a Microsoft Windows component called Local Security Authority
Subsystem Service (LSASS). Like the previous worms, Sasser affected not only corporate networks
but also control networks, which it reached because they were connected to the Internet. One case was
a chemical plant in the USA, where Sasser infected the DCS, causing problems in the HMI and losing
control of the plant for 5 hours. Another case, this time in the U.K., was an oil company, where Sasser
infected several HMI in the control network due to non-updated antivirus.
Steel Plant infection with Ahack Worm (Ahmadian et al., 2020). In 2008, a worm infected a steel
plant in Brazil. An employed used a G3 modem to connect to the Internet from the plant. This action
infected the employed computer and the power and blast furnace plants with the Ahack worm, which
started sending data packets over the network at high speed, flooding the network. This caused a loss of
communication between PLC and SCADA and, therefore, problems in several HMI. The worm caused a
loss of production due to many SCADA servers being rebooted, which subsequently caused financial loss.
Stuxnet (Lagner, 2011). In 2010, a new malware was discovered in a uranium enrichment plant of
Iran, marking a turning point. The malware was known as Stuxnet. The main Stuxnet feature was its
sophistication. It used up to five zero-day exploits to attack ICS. The main goal of Stuxnet was to repro-
gram PLC in the centrifuges to change its speed. The Stuxnet attack consisted of four steps. The first
step was the intrusion inside the Natanz nuclear plants. The intrusion is believed to have been caused
by an infected USB drive. Someone plugged an infected USB drive into a computer inside the internal
81
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
network, and the worm started to replicate itself. The second step was the spread through the network,
looking for computers that control centrifuges. The goal of centrifuges is to spin material at high speed
to separate their components. In the Natanz nuclear plants, the centrifuges tried to separate the uranium
types to isolate the type used for both nuclear power and weapons. The final step of Stuxnet was to
reprogram the PLC that control the centrifuges to modify their speeds. This was accomplished in two
ways. First, it made the centrifuges spin at a higher rate than the normal state for 15 minutes. After 15
minutes, Stuxnet reprograms the PLC again to back it to normal speed. The second way to modify the
centrifuges speed rate was to slow down for 50 minutes. These two ways were repeated for months until
Stuxnet was discovered.
Over time, the exceeded speed of centrifuges caused that many of them broke. It was reported that
around 20% of centrifuges in the Natanz plants were decommissioned.
Considering the strong interest that many counties have in nuclear power and weapons, Stuxnet was
considered the first weapon in cyber warfare. Although nobody knows the authorship of Stuxnet, it is
believed that State intelligence services are behind the attack. Authorship by groups of hackers without
connection to the government was ruled out due to the complexity of the malware. Furthermore, Stuxnet
targeted particular devices and software, such as PLC Siemens and WinCC SCADA software, that are
prohibitively expensive for an ordinary hacker group.
Flame (Munro, 2012). In 2012 a new malware was discovered in the middle east. It is considered
a cyber-espionage weapon like Stuxnet. After infecting a device, Flame spied on the activity of the
machine and stole data from it with keystroke monitoring and packet sniffing functionality. In addition,
the malware has backdoor capabilities that enable cyber attackers to update the malware and erase it as
desired. Flame was a modular malware, and it consisted of more than 20 different modules. Although
the first contagion was in 2010, two years passed until it was discovered.
Shamoon attacks Saudi Arabian Oil Company (Zhioua, 2013). In 2012, the Shamoon virus reached
the national oil company of Saudi Arabia, Aramco. This time, the attack had political intentions and was
launched by a hacker group called Cutting Sword of Justice. The virus reached the internal corporate
network and erased the hard drive of about 30,000 computers. To clean up the malware, Aramco was
forced to disconnect its internal network for more than a week.
Dragonfly targets their attacks to Energy Industry (Kshetri et al., 2017). Dragonfly is a hacker group
whose activity dates back to 2011. After a period of inactivity, they returned to activity in 2014, target-
ing their attacks on the energy industry. From 2014 to nowadays, they have been the authors of several
cyberattacks, such as those targeting Ukraine’s power system in 2015 and 2016. Dragonfly uses different
vectors of attack to access the network victim. Among them, we can find malicious emails, watering
hole attacks, and Trojanized software.
Wannacry (Ehrenfeld, 2017). In 2017, Wannacry infected near 300,000 computers around the world.
Although Wannacry was not aimed to attack ICS, it reached many critical infrastructure systems such
as hospitals and manufacturer factories. Wannacry is a ransomware that encrypts files in the computer
hard drive and asks for a ransom in bitcoins. It is based on an exploit called EternalBlue, stolen by The
Shadow Brokers hacker group from the U.S. National Security Agency (NSA). EternalBlue exploits a
vulnerability in Microsoft Samba protocol, allowing the ransomware to spread itself across the network.
EKANS, a ransomware specially designed to attack ICS (Rege et al., 2020). It can force computers
to stop specific processes. This is performed by killing the applications that control the processes. It
is accomplished by a static kill list stored in the EKANS source code. Once the processes are stopped,
82
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
and the hard drive encryption is finished, a ransom note is stored in the root of the system drive and on
the user’s desktop.
Figure 3 summarizes the most important cyberattacks over the years.
There are several datasets focused on traditional networks to test the performance of ML/DL models
objectively. The most well-known dataset is called KDDCup99 (KDD Cup 99 Dataset), which is an
evolution of the Darpa98 dataset (DARPA 98 Dataset). Although KDDcup99 has several deficiencies
(Tavallaee et al., 2009), it is still widely used in research work in the field of AD in traditional networks.
To solve these deficiencies, a new dataset called NSL-KDD (NSL-KDD Dataset) was developed. An-
other widely used dataset is CTU-13 (Garcia et al., 2014), which is a botnet-oriented dataset, and it was
collected from real network traffic at the Czech Technical University. CTU-13 includes malicious traffic
from different bots combined with normal and background network traffic. In contrast, CICIDS 2017
(Sharafaldin et al., 2018) is a dataset collected from a simulated scenario that contains traffic from two
different networks: the outside and inside network. To carry out the network traffic capture, 4 computers
were infected, launching cyberattacks against 14 victim computers.
Unlike traditional networks, the number of industrial-oriented datasets is quite limited. In the follow-
ing lines, we summarize the most relevant works found in the literature concerning industrial datasets.
In (Industrial Control System (ICS) Cyber Attack Datasets), there are three available industrial da-
tasets. The Power System Dataset (Pan et al., 2015) was collected from a power system, while several
false data injection attacks were executed. It includes logs from the Snort IDS and measurements from
Synchrophasor. The Gas Pipeline Dataset (Beaver et al., 2013) was collected from a gas pipeline while
both false data injection and Denial-of-Service (DoS) attacks were executed. The dataset includes dif-
ferent features from the PLC that controls the process. Finally, Water Storage Dataset (Morris et al.,
2011) was collected from a water tank on which three different attacks were launched: reconnaissance,
false data injection, and DoS attacks (Morris et al., 2011).
Another set of industrial-oriented datasets were developed by the Center for Cybersecurity Research,
iTrust (Centre for Research in Cyber Security, iTrust). All the datasets developed were collected from real
testbeds in their facilities, while false data injection attacks were executed. In particular, we highlight the
83
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
following datasets: Secure Water Treatment (SwaT) (Mathur et al., 2016), Water Distribution (WADI)
(Water Distribution (WADI) Dataset), and Electric Power and Intelligent Control (EPIC) (Adepu, 2018).
The three datasets contain normal and malicious samples from false injection attacks. All of them are
available upon prior request.
The BATADAL (Taormina et al., 2018) (BATtle of the Attack Detection ALgorithms) competition,
which is carried out to propose algorithms for the detection of cyberattacks affecting industrial environ-
ments, provides participants with a dataset consisting of two training subsets. The first of them does not
contain attacks and was generated from a simulation running for one year. The second training subset
was collected for six months and contains data captured during the operation of the plant under several
false data injection attacks.
Finally, to overcome the lack of a methodology to collect datasets from industrial scenarios, the work
presented in (Perales Gómez et al., 2019) is proposed. The methodology consists of 4 steps: attack se-
lection, attack deployment, traffic capture, and feature computation. In addition, the authors presented
the Electra dataset, which was collected from an electric traction substation widely used in the railway
industry. The dataset comprises two subsets: the first contains features from Modbus TCP network traf-
fic, while the second subset contains features from S7Comm network traffic.
The usage of ML and DL techniques has been widely studied both on traditional networks and IoT
scenarios. For example, the authors of (Cvitić et al., 2020) propose an ensemble ML approach to clas-
sify IoT devices in smart homes regardless of their purpose or functionality. The results showed that
this approach can classify IoT devices into four previously defined classes with an accuracy of 99.79%.
Another example is proposed in (Yin et al., 2020), where authors presented an integrated solution that
uses Convolutional Neural Networks (CNN) and recurrent autoencoder to detect anomalies in IoT de-
vices. In addition, the authors used a two-stage sliding window in the data preprocessing step to extract
high-order features.
Since data from sensors and actuators follow a temporal pattern, anomaly detection in industrial sce-
narios is frequently treated as a time series classification problem. Therefore, many authors developed
techniques to deal with time, especially 1-Dimensional Convolutional Neural Networks (1D CNN) and
Long-Short Term Memory (LSTM) neural networks. For example, in (Kravchik et al., 2018), the authors
analyzed an LSTM and 1D CNN to detect anomalies. The authors also proposed a method to compute
high-order features based on calculating the difference between the current and the past value with a
given lag. The authors of (Shalyga et al., 2018) tested Dense Neural Networks (DNN), 1D CNN, and
LSTM networks as core models of an automatic architecture optimization based on Genetic Algorithms
(GAs). The architecture proposed used the Numenta Anomaly Benchmark (NAB) metric (Lavin et al.,
2015) instead of the widely extended F1-score. The authors of (Zizzo et al., 2019) presented an IDS
based on an LSTM model where Cumulative Sum (CUSUM) was proposed as a method to compute the
threshold to trigger an anomaly. A novel Generative Adversarial Network-based Anomaly Detection
(GAN-AD) method based on LSTM models was presented in (Li et al., 2018).
Other ML/DL models than CNN and LSTM networks have also been proposed in the literature. For
example, the authors of (Inoue et al., 2017) evaluated DNN and SVM to detect anomalies in Cyber-
84
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Physical Systems (CPS). The test dataset used in the evaluation phase was scaled with its own mean
and standard deviation. The usage of AutoEncoders (AE) was explored in (Kravchik et al., 2019). The
authors suggested carrying out both filtering and extraction feature steps. The filtering step was intended
to select those features most suitable to model an anomaly detection task. In comparison, the extraction
process computes features in the frequency domain. However, feature extraction creates features based
on the most energetic bands, removing important information from the remaining signal. In addition, the
threshold function used to determine the presence of an anomaly is defined as a function of the mean
and standard deviation of the test dataset.
Regarding the semi-supervised approach, the authors of (Elnour et al., 2020) proposed a novel method
named Dual Isolation Forest (DIF), where two Isolation Forests (IF) were trained. The original scaled
training data was used to train the first IF. While the second IF was trained with a new training dataset
resulting from applying the Principal Component Analysis (PCA) to the original training dataset. The
training dataset to train both IF contained normal data from normal and test datasets. While the test
dataset only contained data from cyberattacks.
Concerning methodology to train AD-based IDS in industrial scenarios, (Pinelli et al., 2003; Fabio
et al., 2018) presented one of the most prominent works. The authors proposed two methodologies for
gas turbines. The first one presented the k-σ methodology, whereas the second presented an optimization
over the k-σ methodology. A significant limitation of these methodologies is that they are based on the
mean and the standard deviation of the data, detecting only a limited type of cyberattacks.
To solve the lack of an AD-based methodology using ML and DL techniques, the authors of (Perales
Gómez et al., 2020) presented MADICS, a methodology for anomaly detection in industrial scenarios. The
methodology consists of five steps: dataset preprocessing, feature filtering, feature extraction, anomaly
detection, and validation. In addition, the authors implemented MADICS in the SWaT testbed, which is
a scaled-down but fully operational raw water purification plant. The authors achieved a state-of-the-art
precision score (0.984), while both the recall (0.750) and the F1-score (0.851) metrics were above the
average of relevant works in the literature. Table 2 shows a comparative analysis of the different ML/
DL techniques examined.
However, it is worth mentioning the vulnerability of ML and DL models to evasion attacks (Perales
Gómez et al., 2021). These attacks consist in generating adversarial samples in the IDS inference phase
which will be misclassified by the IDS, eventually reaching the target industrial device and, therefore,
impacting the physical world, potentially damaging workers and assets.
The solutions presented in Section 4 only consider the cybersecurity aspect, but attacks shown in Section
2 not only impacted the cyber layer but also on the physical layer causing the degradation of services and
potentially injuring citizens and workers. As an example, we show a specific attack detailed graphically
in Figure 4. This attack is carried out in a factory with a production line composed of a pool of conveyor
belts that transport raw products. In addition, several six-axis robotic arms perform specific actions on
those raw products. Each of the conveyor belts is controlled by a PLC that receives commands from the
SCADA server following a Master/Slave architecture. That is, the SCADA initiates the communication,
whereas the PLC can only respond to the SCADA. In addition, the SCADA server has an HMI where
the status of the controlled process is shown.
85
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Due to a vulnerability in a computer of the corporate network, a bot is introduced. From the corporate
network, the bot compromises several legacy devices in the control network forming a botnet. Each of the
infected devices is capable of capturing and modifying the traffic network. First, the operator requests
that the arm robot base be rotated by 15 degrees (Step 1). This action is sent to the SCADA server that
is in charge of controlling and supervising the process. In turn, the SCADA server sends the control
command to the PLC in charge of controlling the specific arm robot. However, a device controlled by
the bot intercepts the command (Step 2) and modifies it to change the rotation angle to 180 degrees (Step
3). Then, the PLC receives the command and rotates the base motor of the robot arm by 180 degrees
(Step 4). Currently, the base motor is already rotated by 180 degrees and may injure workers who share
space with the robot. Although the cyberattack that puts at risk the safety of workers is already carried
out, the bot can modify the response from the arm robot to the HMI to hide its presence. In this case,
the response is sent from the arm robot to the PLC (step 5), and the PLC transmits the response to the
SCADA server (Step 6). At this moment, the device controlled by the bot intercepts the response and
changes the rotation to 15 degrees (Step 7). Finally, the response arrives at the SCADA server, shown in
the HMI (Step 8). Since the bot modified the response of the robot arm, the HMI shows that the action
performed was to rotate the base motor by 15 degrees instead of 180 degrees.
86
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Until now, cybersecurity and safety management in industrial scenarios have been usually carried out by
different and isolated systems. In this context, there are many works focused on safety or cybersecurity.
Cybersecurity solutions have been reviewed in the previous section. Regarding the existence of systems
to improve safety in the industry, the authors of (Vignali et al., 2019) presented a software based on
augmented reality and IIoT intending to increase workers’ safety. The software took data from the server
and sent it to the operators’ mobile devices to report anomalies, allowing rapid response. In addition,
the authors of (Lien et al., 2017) proposed the usage of robots to monitor the environments and detect
obstacles. These robots were equipped with small and low-cost radars. The information collected by the
robots was sent to a safety controller to be studied and the operation state inferred.
However, after studying the works related to cybersecurity and safety integration, we determined
that they are quite limited. To put it in context, the authors of (Lyu et al., 2019) presented a review of
existing works related to risk assessment and management from the cybersecurity and safety perspec-
tive and their integration. In addition, the authors of (Khan et al., 2018) pointed out the importance of
managing cybersecurity and safety in an integrated way in future food manufacturing systems. Among
the limited work available, we highlight the paper presented in (Riel et al., 2017), where authors pro-
posed a method to deal with major challenges of design for safety and cybersecurity in the automotive
context. Similarly, the authors of (Khalid et al., 2017) presented a solution based on addressing issues
in the systematic development of safe and secure human-robot collaboration.
The authors of (Khalid et al., 2018) presented a framework to improve security in spaces where hu-
mans co-work with robots. The proposed solution was based on a strategy form of two components. The
first component allows deploying secure adaptors using cybersecurity methods. The second component
provides calibration support and comparison in real-time from the reference sensor library through an
independent and intelligent module.
Special mention deserves SafeMan, a framework capable of managing cybersecurity and safety in a
unified way (Perales Gómez et al., 2020). SafeMan was proposed as an extensible and flexible framework
made up of different modules and applications deployed as needed. The modules of SafeMan are four,
and each one has a specific responsibility. The Operation Support System (OSS) is in charge of storing
the required policies by other modules and specify their behavior. The Risk Assessment App Manager
(RAAM) deploys the monitoring applications at the edge of the network. Integrated Risk Assessment
System (IRAS) is in charge of analyzing the output of the applications deployed by RAAM to evaluate
the risk. Finally, Action Decision System (ADS) is responsible for carrying out the decision determined
by IRAS.
In addition, SafeMan defines three applications that the RAAM subsystem can deploy. The Cyber
Threat Detection (CTD) application is in charge of analyzing the network traffic to detect cyberattacks.
To this end, ML/DL algorithms for AD to model the regular traffic and detect anomalous activity are
proposed. The Indoor Localization and Activity Recognition (ILAR) application is in charge of tracking
workers’ position and their activities to detect safety threats. To work properly, this application requires
a sensorized environment based on both radio-frequency and computer vision technologies. Finally, the
Safety in Shared-Workspace (SSW) application was proposed to monitor the workspace where human
employees co-work with robots to detect safety threats. As a result, using these applications, SafeMan can
detect potentially dangerous situations, such as the presence of workers in restricted areas, the presence of
objects that pose an obstacle in the path of robots, or the detection of robot malfunctions, to name a few.
87
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Our factories have changed from an isolated to an interconnected approach. This has caused that many
factories have been the target of cyberattacks over the years. Most of these cyberattacks have been carried
out by malware, and to put in context the increment of these cyberattacks, the chapter reviewed different
incidents caused by cyberattacks in industrial scenarios.
In addition, in the chapter, we also reviewed the three most important challenges that need to be
solved to implement secure automation in the industry. The first of them states the need for a framework
capable of managing cybersecurity and safety in a unified way. In Section 5, we introduced an example
of a cyberattack causing a safety threat to illustrate this need. This highlights the need to consider cy-
berattacks as a source of a safety threat. Therefore, they need to be managed in a unified way.
The second challenge states the lack of industrial-oriented datasets needed to train an AD system in
an industrial scenario. In Section 3, we have reviewed the most prominent industrial datasets. However,
we highlighted the need for a methodology to generate such datasets because an industrial dataset col-
lected from a water distribution factory cannot be used to train AD systems in power generation factories.
The final challenge states the need for a methodology to train AD systems in industrial scenarios.
In Section 4, we reviewed the current solutions using ML/DL in the industrial context to highlight this
point. Most of these works use ML/DL models capable of working with time-series to detect anomalies.
However, each author applies different steps without standard criteria. This can lead to cause method-
ological errors in their solutions.
Finally, we showed three solutions that solve the aforementioned challenges. Electra dataset and its
associate methodology to generate industrial-oriented datasets overcome the first challenge. MADICS,
which is a methodology for AD in industrial scenarios, solves the second challenge. Finally, the Safe-
Man framework, which manages cybersecurity and safety in a unified way, solves the third challenge
discussed in the chapter since it considers cyberattacks as a source of a safety threat.
As future work, we plan several lines of work. One of them is applying MADICS methodology to
different industrial scenarios that follow the Industry 4.0 paradigm. Another line is the implementation
of SafeMan on a real manufacturing plant to manage cybersecurity and safety. In this way, we expect
to detect harmful situations that put at risk the workers’ safety and, therefore, reduce the accidents in
the factory.
ACKNOWLEDGMENT
This work has been funded by Spanish Ministry of Science, Innovation and Universities, State Research
Agency, FEDER funds, under Grant RTI2018-095855-B-I00, and the Swiss Federal Office for Defence
Procurement (armasuisse) (project code and CYD-C-2020003).
REFERENCES
Adepu, S., Kandasamy, N. K., & Mathur, A. (2018). Epic: An electric power testbed for research and
training in cyber physical systems security. In Computer Security (pp. 37–52). Springer.
88
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Ahmadian, M. M., Shajari, M., & Shafiee, M. A. (2020). Industrial control system security taxonomic
framework with application to a comprehensive incidents survey. International Journal of Critical In-
frastructure Protection, 29, 100356. doi:10.1016/j.ijcip.2020.100356
Beaver, J. M., Borges-Hink, R. C., & Buckner, M. A. (2013, December). An evaluation of machine
learning methods to detect malicious SCADA communications. In 2013 12th international conference
on machine learning and applications (Vol. 2, pp. 54-59). IEEE. 10.1109/ICMLA.2013.105
Boyer, S. A. (1999). SCADA: Supervisory control and data acquisition (Vol. 3). ISA.
Boyes, H., Hallaq, B., Cunningham, J., & Watson, T. (2018). The industrial internet of things (IIoT): An
analysis framework. Computers in Industry, 101, 1–12. doi:10.1016/j.compind.2018.04.015
Centre for Research in Cyber Security. (n.d.). iTrust. https://itrust.sutd.edu.sg/
Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: A survey. ACM computing surveys
(CSUR), 41(3), 1-58.
Cvitić, I., Peraković, D., Periša, M., & Gupta, B. (2021). Ensemble machine learning approach for clas-
sification of IoT devices in smart home. International Journal of Machine Learning and Cybernetics,
1-24. https://www.ll.mit.edu/r-d/datasets/1998-darpa-intrusion-detection-evaluation-dataset
DatasetN. S. L.-K. D. D. (n.d.). https://www.unb.ca/cic/datasets/nsl.html
Ehrenfeld, J. M. (2017). Wannacry, cybersecurity and health information technology: A time to act.
Journal of Medical Systems, 41(7), 104. doi:10.100710916-017-0752-1 PMID:28540616
Elnour, M., Meskin, N., Khan, K., & Jain, R. (2020). A dual-isolation-forests-based attack detection
framework for industrial control systems. IEEE Access: Practical Innovations, Open Solutions, 8,
36639–36651. doi:10.1109/ACCESS.2020.2975066
Fabio Ceschini, G., Gatta, N., Venturini, M., Hubauer, T., & Murarasu, A. (2018). Optimization of sta-
tistical methodologies for anomaly detection in gas turbine dynamic time series. Journal of Engineering
for Gas Turbines and Power, 140(3), 032401. doi:10.1115/1.4037963
Garcia, S., Grill, M., Stiborek, J., & Zunino, A. (2014). An empirical comparison of botnet detection
methods. Computers & Security, 45, 100-123.
Hastie, T., Tibshirani, R., & Friedman, J. (2009). The elements of statistical learning: data mining,
inference, and prediction. Springer Science & Business Media. Industrial Control System (ICS) Cyber
Attack Datasets. https://sites.google.com/a/uah.edu/tommy-morris-uah/ics-data-sets
Inoue, J., Yamagata, Y., Chen, Y., Poskitt, C. M., & Sun, J. (2017, November). Anomaly detection for
a water treatment system using unsupervised machine learning. In 2017 IEEE International Confer-
ence on Data Mining Workshops (ICDMW) (pp. 1058-1065). IEEE. http://kdd.ics.uci.edu/databases/
kddcup99/kddcup99.html
Khalid, A., Kirisci, P., Ghrairi, Z., Thoben, K. D., & Pannek, J. (2017, July). Towards implementing
safety and security concepts for human-robot collaboration in the context of Industry 4.0. In 39th Inter-
national MATADOR Conference on Advanced Manufacturing (Manchester, UK).
89
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Khalid, A., Kirisci, P., Khan, Z. H., Ghrairi, Z., Thoben, K. D., & Pannek, J. (2018). Security frame-
work for industrial collaborative robotic cyber-physical systems. Computers in Industry, 97, 132–145.
doi:10.1016/j.compind.2018.02.009
Khan, Z. H., Khalid, A., & Iqbal, J. (2018). Towards realizing robotic potential in future intelligent food
manufacturing systems. Innovative Food Science & Emerging Technologies, 48, 11–24. doi:10.1016/j.
ifset.2018.05.011
Kravchik, M., & Shabtai, A. (2018, January). Detecting cyber attacks in industrial control systems us-
ing convolutional neural networks. In Proceedings of the 2018 Workshop on Cyber-Physical Systems
Security and PrivaCy (pp. 72-83). 10.1145/3264888.3264896
Kravchik, M., & Shabtai, A. (2021). Efficient cyber attack detection in industrial control systems using
lightweight neural networks and PCA. IEEE Transactions on Dependable and Secure Computing, 1.
doi:10.1109/TDSC.2021.3050101
Kshetri, N., & Voas, J. (2017). Hacking power grids: A current problem. Computer, 50(12), 91–95.
doi:10.1109/MC.2017.4451203
Langner, R. (2011). Stuxnet: Dissecting a cyberwarfare weapon. IEEE Security and Privacy, 9(3), 49–51.
doi:10.1109/MSP.2011.67
Lasi, H., Fettke, P., Kemper, H. G., Feld, T., & Hoffmann, M. (2014). Industry 4.0. Business & Informa-
tion Systems Engineering, 6(4), 239–242. doi:10.100712599-014-0334-4
Lavin, A., & Ahmad, S. (2015, December). Evaluating Real-Time anomaly detection algorithms--The
Numenta anomaly benchmark. In 2015 IEEE 14th International Conference on Machine Learning and
Applications (ICMLA) (pp. 38-44). IEEE.
Li, D., Chen, D., Jin, B., Shi, L., Goh, J., & Ng, S. K. (2019, September). MAD-GAN: Multivariate
anomaly detection for time series data with generative adversarial networks. In International Conference
on Artificial Neural Networks (pp. 703-716). Springer. 10.1007/978-3-030-30490-4_56
Li, P., Salour, M., & Su, X. (2008). A survey of internet worm detection and containment. IEEE Com-
munications Surveys and Tutorials, 10(1), 20–35. doi:10.1109/COMST.2008.4483668
Liao, H. J., Lin, C. H. R., Lin, Y. C., & Tung, K. Y. (2013). Intrusion detection system: A comprehensive
review. Journal of Network and Computer Applications, 36(1), 16–24. doi:10.1016/j.jnca.2012.09.004
Lien, J., Amihood, P. M., Javidan, A. J., Karagozler, M. E., Olson, E. M., & Poupyrev, I. (2017). Embed-
ding Radars in Robots for Safety and Obstacle Detection. Academic Press.
Lyu, X., Ding, Y., & Yang, S. H. (2019). Safety and security risk assessment in cyber-physical systems.
IET Cyber-Physical Systems. Theory & Applications, 4(3), 221–232.
Mathur, A. P., & Tippenhauer, N. O. (2016, April). SWaT: a water treatment testbed for research and
training on ICS security. In 2016 international workshop on cyber-physical systems for smart water
networks (CySWater) (pp. 31-36). IEEE.
90
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Miller, B., & Rowe, D. (2012, October). A survey SCADA of and critical infrastructure incidents.
In Proceedings of the 1st Annual conference on Research in information technology (pp. 51-56).
10.1145/2380790.2380805
Mirian, A., Ma, Z., Adrian, D., Tischer, M., Chuenchujit, T., & Yardley, T. (2016, December). An internet-
wide view of ICS devices. In 2016 14th Annual Conference on Privacy, Security and Trust (PST) (pp.
96-103). IEEE. 10.1109/PST.2016.7906943
Moore, D., Paxson, V., Savage, S., Shannon, C., Staniford, S., & Weaver, N. (2003). The spread of the
sapphire/slammer worm. CAIDA, ICSI, Silicon Defense, UC Berkeley EECS and UC San Diego CSE.
Morris, T., & Gao, W. (2014, March). Industrial control system traffic data sets for intrusion detec-
tion research. In International Conference on Critical Infrastructure Protection (pp. 65-78). Springer.
10.1007/978-3-662-45355-1_5
Morris, T., Srivastava, A., Reaves, B., Gao, W., Pavurapu, K., & Reddi, R. (2011). A control system
testbed to validate critical infrastructure protection concepts. International Journal of Critical Infra-
structure Protection, 4(2), 88–103. doi:10.1016/j.ijcip.2011.06.005
Munro, K. (2012). Deconstructing flame: The limitations of traditional defences. Computer Fraud &
Security, 2012(10), 8–11. doi:10.1016/S1361-3723(12)70102-1
Nicholson, A., Webber, S., Dyer, S., Patel, T., & Janicke, H. (2012). SCADA security in the light of
Cyber-Warfare. Computers & Security, 31(4), 418–436. doi:10.1016/j.cose.2012.02.009
Pan, S., Morris, T., & Adhikari, U. (2015). Developing a hybrid intrusion detection system using data mining
for power systems. IEEE Transactions on Smart Grid, 6(6), 3104–3113. doi:10.1109/TSG.2015.2409775
Perakovic, D., Perisa, M., Cvitic, I., & Zoric, P. (2020). Identification of the relevant parameters for
modeling the ecosystem elements in Industry 4.0. In 4th EAI International Conference on Management
of Manufacturing Systems (pp. 111-123). Springer. 10.1007/978-3-030-34272-2_11
Peraković, D., Periša, M., Cvitić, I., & Zorić, P. (n.d.). Information and communication technologies for
the society 5.0 environment. Academic Press.
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdrán, A., & García Clemente, F. J. (2020).
MADICS: A Methodology for Anomaly Detection in Industrial Control Systems. Symmetry, 12(10),
1583. doi:10.3390ym12101583
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdran, A., García Clemente, F. J., Cadenas
Sarmiento, C., Del Canto Masa, C. J., & Méndez Nistal, R. (2019). On the generation of anomaly de-
tection datasets in industrial control systems. IEEE Access: Practical Innovations, Open Solutions, 7,
177460–177473. doi:10.1109/ACCESS.2019.2958284
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdrán, A., García Clemente, F. J., & Cleary, F.
(2021). Crafting Adversarial Samples for Anomaly Detectors in Industrial Control Systems. In The 4th
International Conference on Emerging Data and Industry 4.0 (EDI40).
91
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdrán, A., García Clemente, F. J., Gil Pérez,
M., & Martínez Pérez, G. (2020). SafeMan: A unified framework to manage cybersecurity and safety
in manufacturing industry. Software, Practice & Experience.
Pillitteri, V. Y., & Brewer, T. L. (2014). Guidelines for smart grid cybersecurity. Academic Press.
Pinelli, M., Venturini, M., & Burgio, M. (2003, January). Statistical methodologies for reliability as-
sessment of gas turbine measurements. In Turbo Expo: Power for Land, Sea, and Air (Vol. 36851, pp.
787-793). 10.1115/GT2003-38407
Rege, A., & Bleiman, R. (2020, June). Ransomware Attacks Against Critical Infrastructure. In ECCWS
2020 20th European Conference on Cyber Warfare and Security (p. 324). Academic Conferences and
Publishing Limited. https://www.risidata.com/
Riel, A., Kreiner, C., Macher, G., & Messnarz, R. (2017). Integrated design for tackling safety and security
challenges of smart products and digital manufacturing. CIRP Annals, 66(1), 177–180. doi:10.1016/j.
cirp.2017.04.037
Shalyga, D., Filonov, P., & Lavrentyev, A. (2018). Anomaly detection for water treatment system based
on neural network with automatic architecture optimization. arXiv preprint arXiv:1807.07282.
Sharafaldin, I., Lashkari, A. H., & Ghorbani, A. A. (2018, January). Toward generating a new intrusion detec-
tion dataset and intrusion traffic characterization. In ICISSp (pp. 108-116). doi:10.5220/0006639801080116
Skobelev, P. O., & Borovik, S. Y. (2017). On the way from Industry 4.0 to Industry 5.0: from digital
manufacturing to digital society. Industry 4.0, 2(6), 307-311.
Taormina, R., Galelli, S., Tippenhauer, N. O., Salomons, E., Ostfeld, A., Eliades, D. G., Aghashahi, M.,
Sundararajan, R., Pourahmadi, M., Banks, M. K., Brentan, B. M., Campbell, E., Lima, G., Manzi, D.,
Ayala-Cabrera, D., Herrera, M., Montalvo, I., Izquierdo, J., Luvizotto, E. Jr, ... Ohar, Z. (2018). Battle of
the attack detection algorithms: Disclosing cyber attacks on water distribution networks. Journal of Water
Resources Planning and Management, 144(8), 04018048. doi:10.1061/(ASCE)WR.1943-5452.0000969
Tavallaee, M., Bagheri, E., Lu, W., & Ghorbani, A. A. (2009, July). A detailed analysis of the KDD
CUP 99 data set. In 2009 IEEE symposium on computational intelligence for security and defense ap-
plications (pp. 1-6). IEEE.
Vignali, G., Bottani, E., Guareschi, N., Di Donato, L., Ferraro, A., & Pirozzi, M. (2019, June). Develop-
ment of a 4.0 industry application for increasing occupational safety: guidelines for a correct approach.
In 2019 IEEE International Conference on Engineering, Technology and Innovation (ICE/ITMC) (pp.
1-6). IEEE. 10.1109/ICE.2019.8792814
Vinayakumar, R., Alazab, M., Soman, K. P., Poornachandran, P., Al-Nemrat, A., & Venkatraman, S.
(2019). Deep learning approach for intelligent intrusion detection system. IEEE Access: Practical In-
novations, Open Solutions, 7, 41525–41550. doi:10.1109/ACCESS.2019.2895334
Water Distribution (WADI) Dataset. (n.d.). https://itrust.sutd.edu.sg/testbeds/water-distribution-wadi/
Wu, H., Schwab, S., & Peckham, R. L. (2008). U.S. Patent No. 7,424,744. Washington, DC: U.S. Patent
and Trademark Office.
92
Malware Detection in Industrial Scenarios Using Machine Learning and Deep Learning Techniques
Xin, Y., Kong, L., Liu, Z., Chen, Y., Li, Y., Zhu, H., Gao, M., Hou, H., & Wang, C. (2018). Machine
learning and deep learning methods for cybersecurity. IEEE Access: Practical Innovations, Open Solu-
tions, 6, 35365–35381. doi:10.1109/ACCESS.2018.2836950
Yin, C., Zhang, S., Wang, J., & Xiong, N. N. (2020). Anomaly detection based on convolutional recur-
rent autoencoder for IoT time series. IEEE Transactions on Systems, Man, and Cybernetics. Systems,
1–11. doi:10.1109/TSMC.2020.2968516
Zhioua, S. (2013, July). The middle east under malware attack dissecting cyber weapons. In 2013 IEEE
33rd International Conference on Distributed Computing Systems Workshops (pp. 11-16). IEEE. 10.1109/
ICDCSW.2013.30
Zizzo, G., Hankin, C., Maffeis, S., & Jones, K. (2019). Intrusion detection for industrial control systems:
Evaluation analysis and adversarial attacks. arXiv preprint arXiv:1911.04278.
93
94
Chapter 6
Malicious Node Detection
Using Convolution Technique:
Authentication in Wireless
Sensor Networks (WSN)
Priyanka Ahlawat
National Institute of Technology, Kurukshetra, India
Pranjil Singhal
National Institute of Technology, Kurukshetra, India
Khushi Goyal
National Institute of Technology, Kurukshetra, India
Kanak Yadav
National Institute of Technology, Kurukshetra, India
Rohit Bathla
National Institute of Technology, Kurukshetra, India
ABSTRACT
Considering the situation when the node which is entering in the network is a malicious node, it can
check and record the various operations of the network which could be responsible for some serious
security issues. In order to safeguard the network, the author propose an approach which is in view of
the advanced encoding strategy, termed as convolution coding approach. According to the requirement
of network, the initial bits are assigned, and by applying the convolution technique, the final code is
generated for each and every node. Considering the fact that it is a digital method, the codes can be
represented by using the binary number system. All the nodes that are a part of the network will be having
their corresponding final binary code, say C. The verification of each node is carried out by matching
the generated code C with the security code within a particular time period before transmitting the data.
This process enables the detection of malicious or attacker node. Furthermore, to enhance the versatility
and execution, the system is organized into clusters.
DOI: 10.4018/978-1-7998-7789-9.ch006
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Malicious Node Detection Using Convolution Technique
INTRODUCTION
Wireless sensor networks (WSNs) are basically described in the form of a collection of sensor hubs or
nodes which are spread in a region and utilized for checking and recording various observations, per-
ceptions and states of being of a specific domain and conditions of any particular surrounding. Because
of the autonomous structures of Sensors networks, the nodes are having the permission to enter or then
leave the system or formed network at any instant of time. It makes the system vulnerable and prone to
attacks. The main characteristics of WSNs are their infrastructure-less nature and the ability that they
can be self-configured without the need of any predefined structure or configuration. WSNs comprises
of a set of a number of small senor nodes. The communication carried out in such network is wireless.
Primarily, WSNs were developed for some complex military communications in certain areas where it
was not feasible to establish wired networks. Nowadays wireless sensors are useful in numerous other
applications, considering few as, industrial modelling, logistics, defence and security, forest fire monitor-
ing, monitoring of poisonous gases etc. The key features of a WSN are that they are spatially distributed,
infrastructure less, self organized, do not require any predefined structure. Generally, the sensor nodes
are having very low storage capacity, low computational capabilities and limited supply of energy. Sen-
sor nodes are allowed to be a part or leave the system at any given moment. This autonomous behaviour
of the network which allows node mobility is solely responsible for some serious security concerns.
Any malicious or attacker node can enter into the legitimate network and cause some serious damage.
The secrecy and integrity of the information transferred within the network could be tampered or lost,
threatening the entire sensor network.
Hence, there is a serious need of some security mechanism in order to keep the authenticity, integrity
and confidentiality of the data which is being monitored, collected by sensor nodes and further transferred
in the network for the purpose of its storage, processing and analysis. As these networks could be easily
compromised due to its limitations, it is really critical to complete the task of detection and isolation of
the malicious nodes for avoiding any chance of some further damage (Perrig A.,2002)
Till now a number of approaches are proposed and implemented in order to fulfil the objective of
providing security standards to the network. One of approach among them is cryptography, which basi-
cally executes the task of securing the original data by the use of a key. Encryption could be done by two
methods: symmetric cryptography and asymmetric cryptography. In the case of symmetric cryptography,
we can use the same key for both of the operations .While with later cryptographic technique; we need
to use unique keys for encryption and decryption. As there is a need of keys for encryption and decryp-
tion, each node in the network is to be provided with unique keys. Due to limited supply of energy and
storage space in WSNs it becomes difficult to perform cryptography. Therefore, for the systems having
bigger sets of sensor hubs or nodes, the process of generation of keys becomes cumbersome and more
complex .Later some hybrid cryptographic techniques were also suggested such as a combination which
make use of both type of cryptographic techniques and digital signature. However, the efficiency of these
techniques was also not sufficient enough because the complexity used to increase with the networks
having bigger sets of nodes, we are considering.
While with the implementation of this technique, we are considering to obtain a binary code word
for security purpose, which need to be matched by all the nodes before accessing the information. The
process of generation of code word is to be done into two steps.
Initially we the security bits are assigned according to the network requirement depending upon
its size, which is directly proportional to the number of sensor nodes present in the system. After that
95
Malicious Node Detection Using Convolution Technique
by implementing the convolution code generation technique the final security binary code is obtained
(Alghamdi,2019) . Before accessing the network there is a need of matching the code word with the
generated code for each node within a given time period. As the malicious nodes will not be able to
perform this task .Furthermore in order to get better performance and improve the results of Convolution
technique the system will be arranged in the form of clusters. Here, clustering proves to be a significant
improvement. As the process of matching the code word becomes easy and the scalability of the network
also improves. Hence malicious nodes will be detected before they could tamper the network by using
convolution technique.
Overview
As the nodes of a wireless sensor network (WSN) are equipped in remote areas and vulnerable environ-
ment. The network could be easily compromised by some attackers due to the constraints of a WSN such
as autonomous nature, limited battery lifetime, memory space and computing capability. This could lead
to the reduction in reliability of the data which is to be retrieved and transferred causing some serious
damage. In order to prevent the network from some serious security issues and information breaches
there is a critical need of some effective security measures for our network.
• Isolation of the malicious nodes for avoiding any chance of some further damage
• Mechanism in order to keep the authenticity, integrity and confidentiality of the data which is be-
ing monitored, collected by sensor nodes
• Transferred in the network for the purpose of its storage, processing and analysis.
• Nodes will be detected before they could tamper the network
Organization
The techniques which were proposed earlier were not efficient enough to perform the desired task
.Whereas the convolution technique is really effective enough in fulfilling the security requirements
considering various storage and energy constraints of a wireless network. There is a restriction for all
the nodes to match the code word with the generated code before accessing the network .So that we can
prevent the network and the information being transferred from being compromised. The detection and
elimination of malicious node could be successfully executed.
Hence, the network will be secured from vulnerable attacks and security threats and the information will
be reliable as it will be prevented from various attacks including eavesdropping and cannot be tampered.
LITERATURE SURVEY
Mostly wireless sensor networks are worked without any human supervision and it can easily lead to the
compromise of an individual node. Many approaches have been proposed regarding the security of these
networks considering different aspects of WSN like limited memory, battery lifetime and computational
96
Malicious Node Detection Using Convolution Technique
capacity. Here are some works closely related to identification and isolation of t he malicious node in
the system of sensor nodes.
The combination of two security protocols has been proposed by (Perrig A.,2002) to enhance the
security standards of the system of sensor networks & improve the performance of overall wireless net-
work but it came up with the obvious effects of merging the demerits of both the combined protocols.
The proposed work of Necla and Ismail, 2012reports the use of algorithms of Encryption which is
Scalable enough and another one which is Cipher Block Chaining- Message Authentication Code so that
the security can be ensured in the network on a dynamic level of system. The problem faced with this
approach was that with the increment in the amount of bits to be transmitted, there is a clear increment
in the consumption of power in the network. For the security of mobile ad-hoc networks, a zone routing
model has been proposed by Ranjeetha et al.,2017which uses key distribution. But because of lack of
the central authority, key distribution can be a complex task in such networks.
The approach by Zang et al., 2013 proposes that keys should be maintained in hierarchical networks
which can be done by giving unique keys to the base stations, heads of clusters instead of providing the
same key to all. Zheng et al.,2017gave the method where dynamic trust can be used to provide security.
With the help of direct and indirect trust, dynamic trust can be found and a sliding window is used to
improve flexibility. But, regular update of parameters can be problematic. Noor et al., 2012used the
secondary routing and secondary gateway in cluster network to develop a method which is efficient in
terms of power consumption to provide security to the system. But in autonomous structure of networks,
clustering might be bit difficult. An energy efficient optimized algorithm which improve performance,
improve the reliability of the system and safeguard the network from various forms of attack.,2018
Using node to node interactions, Kaur et al.,2016 proposed a key distribution method which was
secured enough and differentiated attacking node so that a safe routing is mapped is from one initiating
node for communication to the destination. But this method tend to fail in the condition where at any
time interval, a group of malicious nodes enter the network and tends to present themselves as active
nodes. Imad et al.,2016 presented a method in which on considering the value of trust factor of nodes,
trust among the nodes establishes in the system. This trust is maintained in the observations of the nodes
and this trust is responsible for the successful transmission of information. However, each time during
the process of transferring information with trust participating as router, node energy is reduced.
Chan et al.,2003 introduced a technique which used pair-wise key distribution of random key within
the nodes of the system so that reliability in communications can be maintained. But the process of
distribution of the keys needed prior the validation of trust to which node that particular key has been
given. Regarding the topology of the network, a hierarchical network is proposed in as WSN is an energy
sensitive network and this approach can help in energy saving. MECH (maximum energy cluster head)
protocol is a LEACH (low energy adaptive clustering hierarchy) like method where network is divided
into various groups of clusters where every cluster is having a unique cluster head.
Various existing methods use the technique of key distribution, as it consumes more energy, hence
these methods are not efficient enough and needs more storage to obtain and store the encryption and
decryption pin. For more complex and large systems, it becomes difficult to provide pin to each node
when they enter the network. More space for storage is to be provided for that which is generally less.
In the proposed method, an encoding approach is implemented which is already present in the systems
for digital communication. Different code word is obtained at different hop without the help of any ad-
ditional hardware. Key distribution complexity is removed and because of the use of ex-or gates, power
consumption is also low. Thus, the proposed work is suitable for heterogeneous multi-hop networks.
97
Malicious Node Detection Using Convolution Technique
PROPOSED WORK
Network Model
WSN is an energy sensitive model. Thus, we use the MECH (Maximum Energy Cluster Head) protocol
to build the network. It is same as a protocol such LEACH where the network is divided in clusters.
The below Figure 1 shows the architecture of MECH protocol where sensor nodes play two roles:
member node and cluster head. Sensors organize themselves into clusters and one cluster head node is
associated with each cluster, which manages the data between the base station and member nodes and
it also manages the cluster. The main aim of the network model is to enhance the execution of Convo-
lution Technique. Clustering basically helps for improving the scalability of the network. It makes the
network less complex and more organized easing the process of validation of nodes by matching their
respective code words
Code Generation
In this approach, Convolution Coding Approach (a digital encoding technique) creates the security
code. Initially, on the basis of network requirement, security bits can be selected and then in order to
make a security code, convolutional technique is applied on the ISB (Initial Security Bits) and modulo
2 operation is used for that.
All the data, including the security information is broadcasted by the source node to all of its neigh-
bouring nodes, and the node can access the data if it is a legitimate node and then it keeps on broadcasting
the data to its neighbours. And it continues till the information reaches the destination node. If it is not
98
Malicious Node Detection Using Convolution Technique
legitimate, it cannot retrieve the data as it does not have any information about the security operation and
thus cannot produce a security code (CB) in a limited time. In this manner we will be able to identify
and isolate the malicious nodes present in our legitimate network before it can access to the data which
is being shared and transferred in the network.
We propose an approach of Malicious Node Detection using Convolution Technique based on a system-
wide quantitative data flow model. The basic model of our data flows analysis on this incremental and
aggregated approach towards the quantitative data flow graphs. These basic graphs describe the interac-
tion among various system entities such as methods, support, data, or system registries. We demonstrate
the feasibility of our approach through these data flow diagrams.
Level 0 DFD
It’s designed such as abstraction design, explaining the operation as a separate rule with its connection to
outside entities. It describes the whole system as private, including information and output data indicated
by arrows. Our basic approach considering the input to the network that the situation when the node which
is entering in the network is an a malicious node then it can check and record the various operations.
Level 1 DFD
The structure consists of a collection of sensor nodes that will be transmitting the information. This ac-
cess point will regularly receive sensor nodes’ information to collect data sent by sensor nodes. There
are two sorts of attackers, static one and dynamic one. Static, which regularly transfers false informa-
tion to other node & reverse the data whenever is demanded. The dynamic which conveys both accurate
data and wrong data to nodes. Hence the arrangement becomes unclear whether it is a real sensor or an
invalid sensor. Ultimately, we suggest how malicious node can be identified by using convolutional code
generator. In order to safeguard the network we are hereby proposing an approach which is in view of
the advanced encoding strategy, termed as Convolution coding approach.
99
Malicious Node Detection Using Convolution Technique
Level 2 DFD
According to the requirement of network, the initial bits are assigned and by applying the convolution
technique, the final code is generated for each and every node. Considering the fact that it is a digital
method, the codes can be represented by using the binary number system. All the nodes which are a part
of the network will be having their corresponding final binary code. The verification of each node is
carried out by matching the generated code with the security code within a particular time period before
transmitting the data. This process enables the detection of malicious or attacker node. Furthermore to
enhance the versatility and execution, the system is organised into clusters.
100
Malicious Node Detection Using Convolution Technique
101
Malicious Node Detection Using Convolution Technique
IMPLEMENTATION
The implementation of the Convolution Technique for malicious node detection is WSN is done in
Network Simulator 2 (NS2).
What is NS2?
NS2 is an event-driven simulator. To study the dynamic nature of networks, NS2 is proved to be useful.
The various simulations on wired & wireless network protocols can be performed using ns2. The users
can define the network protocols and can simulate the corresponding behaviours using NS2.
• To run the Tcl simulation script file, execute the command “ns” having name of Tcl simulation
script file in the one input argument.
• NS2 contains the two main languages which are: OTcl which stands for (Object-Oriented Tool
Command Language). and C++.
• The backend (i.e. inside mechanism) is defined in C++.
• The frontend (i.e. scheduling discrete events as well as assembling and configuring of objects) is
defined in OTcl.
• TclCl = Tcl + Classes Libraries.
• On execution of Tcl simulation file, a text based simulation trace file is generated.
• To analyse the results interactively, NAM (Network Animator) is used.
• To analyse the results graphically, XGraph is used.
102
Malicious Node Detection Using Convolution Technique
• The hierarchical network is implemented MECH protocol. The basic structure is as follows:
• Each node is provided with a different ID.
• Every sensor node is known and have a fixed position.
• Base station consists of infinite energy.
• All the sensors are fixed in the transmission area of base station.
• The sensor nodes are able to adjust the transmit power.
Cluster Formation
• Base station construct the remaining energy matrix and distance matrix.
• Each and every sensor node sends its location and energy to their base station.
• A network is divided by the base station in k number of clusters as follows:
• Base station construct the remaining energy matrix and distance matrix.
• The matrices are broadcasted to each node by the base station.
• The sensor node having maximum energy in each subdivision is chosen to be a cluster head (CH)
and if the two sensor nodes have same energy, the the node having smallest ID is chosen to be the
cluster head.
• The sink node broadcasts the Cluster Head’s ID to the sensor nodes.
• By using distance matrix, all the sensor nodes associate themselves to the closest Cluster Head.
Re-Clustering
• The Cluster Heads are chosen again if the energy of a Cluster Head becomes less than its threshold
energy.
• Then, the base station updates the energy matrix when the cluster head is changed.
In this approach, we used manageable dispersed data convolution solutions that can quickly adjust to
changeable environmental and thus cognitive performance of malicious nodes.
The following advantages are:
Step 1: Based on the requirement of the network, the initial security bits are selected at each hop. It is
expressed as Isb.
ISB = Hc – 1 (1)
103
Malicious Node Detection Using Convolution Technique
IS1 = 1 -1 = 0
IS2 = 2-1 = 1
IS3 = 3 – 1 = 2
Step 2: The ISB is represented in 3-bits using 3-bit convolutional code generator. It is expressed as:
IS1 = 0 0 0
IS2 = 0 0 1
IS3 = 0 1 0
Step 3: Calculate the bits G1B, G2B, G3B using modulo 2 addition of the ISB (initial security bits) as follows:
104
Malicious Node Detection Using Convolution Technique
G11 = (0 + 0) % 2 = 0
G21 = 0
G31 = 0% 2 = 0
G12 = (0 + 0) % 2 = 0
G22 = 0
G32 = 1
G13 = (0 + 1) % 2 = 1
G23 = 0
G33 = 0% 2 = 0
Step 4: To analyse the trust of the node, obtain the security code (CB) by concatenating the initial se-
curity bits (ISB) with the generated bits (G1B, G2B, G3B) as follows:
N
CB S
i 1
iB GiB
where SiBGiB are put in sequence series and N represents total number of considered initial bits.
For example, At hop 1, i.e. IS1 = 0 0 0, G11 = 0, G21 = 0, G31 = 0, then
CB = S11S21S31G11G21G31 (9)
CB= 0 0 0 0 0 0
CB = S12S22S32G12G22G32 (10)
CB= 0 0 1 0 0 1
CB = S13S23S33G13G23G33 (11)
105
Malicious Node Detection Using Convolution Technique
CB= 0 1 0 1 0 0
Step 5: Verify the CB of every node in the routing process. If the security code is matched in the defined
time, then the node will be considered to be legitimate and can transmit the data otherwise, the
node is considered to be malicious.
Since, the malicious node does not know the security operation, so it takes enough time to generate
the security code, if length of security code is 8 so it generate 28 combinations which is difficult to guess
in the defined TTL.
The proposed approach is applied in simple wireless network as well as hierarchical network to test
the performance of convolution technique.
RESULT
Simulation in NS2
For getting better insights of the performance of the convolution technique, we simulated it using NS2.
The Simulation Parameters are listed below:
Time of simulation: 30 s
Nodes count considered: 3 – 10
Link layer: Logical link
Size of packet: 500 bytes
Propagation model: Two-ray ground
Type of MAC used: 802.11
Network area: 500 m×500 m
Type of queue: Drop-tail
Type of Routing: AODV
Speed of node: 5, 15, 25 m/s.
Initially we just implemented the convolution technique which provided significant simulation results.
Then to improve the performance of this technique, cluster based network architecture is implemented.
The cluster based architecture helped to improve upon the time constraint. For the particular network,
there is an improvement of approximately 4 seconds.
The results of the simulation clearly indicate that the convolution is successful in detection of the
malicious node. It also keeps on focus the energy and storage constraints of the wireless sensor network.
Less computational work is performed as compared to other key distribution mechanism. The performance
is further improved by forming various cluster within the network as it eases the process of validating
legitimate nodes and identification & isolation of malicious nodes present in the network.
106
Malicious Node Detection Using Convolution Technique
107
Malicious Node Detection Using Convolution Technique
108
Malicious Node Detection Using Convolution Technique
Figure 11. Packet drop values for malicious nodes and improved time constraint.
WSNs are having a vast application domain nowadays. They are really compact and efficient in transfer
of data in various physical environments. The approach proposed, through convolutional codes provides
security to wireless networks without using some sort of key distribution, which is not effective because
it consumes more storage space and power. The proposed technique is efficient as we are using simple
mathematical equations for the generation of security code, it thus reduce computational power .Easy
detection of the malicious nodes is possible for secured transmission of information from the source
node to the destination node.
Furthermore, to improve its performance cluster based network architecture was implemented. It re-
sulted in improved value for time constraints. It also increases the scalability of the network. The results
of simulation show that the proposed approach gives quite decent results, and it is easy and helpful to
provide the security to WSN.
REFERENCES
Al-Qerem, A., Alauthman, M., Almomani, A., & Gupta, B. B. (2020). IoT transaction processing
through cooperative concurrency control on fog–cloud computing environment. Soft Computing, 24(8),
5695–5711. doi:10.100700500-019-04220-y
109
Malicious Node Detection Using Convolution Technique
Alghamdi, T. A. (2019). Convolutional technique for enhancing security in wireless sensor networks
against malicious nodes. Human-centric Computing and Information Sciences, 9(1), 38. doi:10.118613673-
019-0198-1
Chen, H., Perrig, A., & Song, D. (2003) Random Key Predistribution Schemes for Sensor Networks.
Proceedings of the 2003 IEEE symposium on security and privacy, 197–213.
Chouhan, L., Chauhan, N., Mahapatra, A. S., & Agarwal, V. (2020). A survey on the applications of
machine learning in wireless sensor networks. International Journal of High Performance Computing
and Networking, 16(4), 197–220. doi:10.1504/IJHPCN.2020.113779
Dwivedi, R. K., Kumar, R., & Buyya, R. (2021). Secure healthcare monitoring sensor cloud with attribute-
based elliptical curve cryptography. International Journal of Cloud Applications and Computing, 11(3),
1–18. doi:10.4018/IJCAC.2021070101
Imad, J., Mohammed, F., Jaroodi, J. A., & Mohamed, N. (2016). TRAS: a trust-based routing protocol
for ad hoc and sensor networks. IEEE 2nd international conference on big data security on cloud, IEEE
international conference on high performance and smart computing, IEEE international conference on
intelligent data and security, 382–387.
Jerbi, W., Guermazi, A., & Trabelsi, H. (2020). A novel secure routing protocol of generation and manage-
ment cryptographic keys for wireless sensor networks deployed in internet of things. International Journal
of High Performance Computing and Networking, 16(2-3), 87–94. doi:10.1504/IJHPCN.2020.112693
Kaur, J., Gill, S. S., & Dhaliwal, B. S. (2016). Secure trust based key management routing framework
for wireless sensor networks. Journal of Engineering (Stevenage, England), 2016, 1–9.
Learning-Based Security Technique for Selective Forwarding Attack in Clustered WSN. (n.d.). Applied
Soft Computing, 108, 107473. doi:10.1016/j.asoc.2021.107473
Necla, B., & Ismail, E. (2012). WSNSec: A scalable data link layer security protocol forWSNs. Ad Hoc
Networks, 10(1), 37–45. doi:10.1016/j.adhoc.2011.04.013
Noor, Z., Jung, L., Alsaadi, F., & Alghamdi, T. (2012). Wireless sensor network (WSN) routing security,
reliability and energy efficiency. J Appl Sci, 12(6), 593–59. doi:10.3923/jas.2012.593.597
Perrig, A., Szewczyk, R., Wen, V., Cullar, D., & Tygar, J. D. (2002). SPINS: Security protocols for sen-
sor networks. Int J Commun Comput Inform, 8(5), 521–534.
Ranjeetha, S., Renuga, N., & Sharmila, R. (2017) Secure zone routing protocol for MANET. International
conference on emerging trends in engineering, science and sustainable technology (ICETSST-2017), 67–76.
Salhi, D. E., Tari, A., & Kechadi, M. T. (2021). Using Clustering for Forensics Analysis on Internet
of Things. International Journal of Software Science and Computational Intelligence, 13(1), 56–71.
doi:10.4018/IJSSCI.2021010104
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based big data secure management in the
fog over a 6G wireless network. IEEE Internet of Things Journal, 8(7), 5164–5171. doi:10.1109/
JIOT.2020.3033131
110
Malicious Node Detection Using Convolution Technique
Tao, Y., Xiangyang, X., Tonghui, L., & Leina, P. (2018). A secure routing of wireless sensor networks based
on trust evaluation model. Procedia Computer Science, 131, 1156–1163. doi:10.1016/j.procs.2018.04.289
Zhang, Y., Wu, C., Cao, J., & Li, X. (2013). A secret sharing-based key management in a hierarchi-
cal wireless sensor network. International Journal of Distributed Sensor Networks, 2013(6), 1–7.
doi:10.1155/2013/406061
Zhang, Wu, & Cao. (2008). A Secret Sharing-Based Key Management in Hierarchical Wireless Sensor
Network. In Securing wireless sensor networks: A survey. IEEE Communications Surveys and Tutorials.
Zhengwang, Y., Wen, T., Song, X., Liu, Z., & Fu, C. (2017). An efficient dynamic trust evaluation model
for wireless sensor networks. Journal of Sensors, 2017, 1–16.
111
112
Chapter 7
Scalable Rekeying Using Linked
LKH Algorithm for Secure
Multicast Communication
Priyanka Ahlawat
National Institute of Technology, Kurukshetra, India
Kanishka Tyagi
National Institute of Technology, Kurukshetra, India
ABSTRACT
In the real scenario, there is a large multicast group where nodes leave and join frequently, and also the
number of nodes leaving and joining is also not proportionate. Hence, scalable rekeying process is an
important issue that needs to be concerned for the secured group communication for dynamic groups.
In basic rekeying scheme, which is based on the logical key hierarchy, the rekeying cost depends on the
logarithm of the size of group for a join or depart request by the user. However, the memory efficiency
of this group rekeying protocol (GREP) is a huge storage overhead over the system. The authors aim
to provide a survey of various group key management schemes and then propose an efficient scalable
solution based on linked LKH and the linked list data structure. Results have shown that the Linked LKH
algorithm has a very low effective cost for rekeying the LKH as compared to the basic LKH algorithm
(i.e., based on the number of new joined and departure requests).
INTRODUCTION
Group Communication among valid users is a powerful Message exchange Model. Group communication
applications include content distribution over large-scale networks, smart wireless networks, software
updates, Video Conferencing, and many more (He Niedermeier & Meer 2013). Key management se-
curity should have basic requirements like authenticity, integrity, and confidentiality. The factors that
affect any group key management process’s performance are scalability, 1 affects n problem, delays
bandwidth for efficient group key distribution. Scalability refers to the network’s ability to handle large
DOI: 10.4018/978-1-7998-7789-9.ch007
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
dynamic groups, i.e., if the number of users becomes large, the key management scheme should handle
without degrading its performance. 1-affects n problem relates with new key generation after a leave
or joins processes during a rekeying. Quality of service should also be maintained during a rekeying
process, such as control packets, storing keys, and delay induced during encryption and decryption. Key
management is very important in group communication restricting access control (Duma, Shahmehri,
lambrix 2003). Key management establishes and maintains the secret keying relationships between
valid parties according to a policy. It includes member identification and authentication. in this regard,
authentication plays a significant role. Once a new member joins, it has to be validated (Zhu, Jajodia
2003). It is also essential to change or update the group key at regular intervals to effectively maintain
a communicated message’s security. Also, key independence has to be properly maintained in which
each key is independent of another key. It means the method of generating a new group key should be
independent of the previous key generation. It also enhances the security of the overall system. Combi-
natorial optimization of group key is given in (Eltoweissy, Heydari, Morales, Sudborough, 2004). Any
Participant can become a part of Group Communication by becoming a group member explicitly. A
group member holds a secretly shared cryptographic Group key and used the same for exchanging the
messages. When a participant becomes a group member, it is required to maintain backward secrecy,
i.e., the participant cannot decipher the messages exchanged before it’s joined.
Similarly, whenever the group member leaves the group, Group communication must maintain the
forward secrecy. After the group leaves, it must prevent the node from accessing the messages. To
achieve the forward secrecy and backward secrecy, group rekeying is performed, which ensures a new
and different key among the group members when a node leaves or joins the group. However, group
rekeying doesn’t play an effective role if we could ensure the group’s structure by defining the members
through the pre-registration of members (Panda, Thool 2016). Group key management can be classified
into three classes: centralized, decentralized, and distributed key management. In centralized schemes,
a single entity generates, distributes, and management of the group key.
Hence a single entity controls the entire group. Minimization of storage, computational power, and
bandwidth are the key challenges in this scheme. In decentralized schemes, the process of management
of group keys is divided among different group members. Hence a single point of failure is not a problem
in this scheme.
In distributed group key management, No key server is explicitly declared. In this scheme, group
members perform the key generation function. It can be contributory or done by individual members.
Maintaining security in every group communication protocol is a critical issue.
The safety goal in a group communication process is to guarantee access only to valid group mem-
bers. The entry and leave of the group members or users are the main reason for modifying the group
key and giving them greater confidence in secure communication, known as re-keying. Since it is a
frequently performed activity during group communication, it is necessary to do the group key update
in a scalable and efficient way. Previously, the client-server paradigm is the most commonly used tech-
nique for conferencing, chat groups, immersive video games, etc., that use the principle of unicast for
data transmission. Present-day developments in Internet technology, with the increase in bandwidth,
are encouraging new developments in the environment. Unlike the old network communication models,
where packets are to be delivered in a unicast model, the multicasting technique provides an effective
delivery service to a larger user community with efficient and effective network resources (Xu, Y., &
Sun, Y 2005). The key tree approach is efficiency depends crucially on whether the key tree stays bal-
anced. Rebalancing with fixed time intervals is used to balance the key tree if it becomes unbalanced.
113
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
This increases the network overhead. With the advancement of many group-oriented applications like
pay-per-view, group-oriented mobile commerce, and military applications, security service models are
required to support group-oriented privacy and data integrity in communication. One realistic solution is
to use a standard group key to encrypt the transmitted messages that can only be accessible to approved
or valid members. As group membership may be dynamic, this group key needs to be securely updated
and redistributed to all authorized or valid group members if there is a change in group membership.
It is used to provide secrecy in the forward direction and backward direction. Forward secrecy means a
member leaving or departing a group may get details about future group messages, and backward secrecy
means a valid group member entering can get information about past group communication. The primary
update method is called the rekeying and rekeying expense, which denotes the number of messages that
need to be disseminated to the members during rekeying. A variety of scalable approaches have been
proposed, and this chapter analyzes the tree approach in detail. In short, the key tree method uses a key
hierarchy, where each member is allocated a set of keys based on their position in the key tree. The key
tree method is rekeying cost increases linearly with the group size logarithm for a request to join or depart.
However, the effectiveness of the key tree approach depends on whether the key tree is balanced or
not. The key tree is balanced if the distance from the root to any two leaf nodes differs by no more than
one. The height from the root to any of the leaf nodes is log k (N) for a balanced key tree with N group
members. K is the out-degree of the key tree, but if the key tree is unbalanced, the distance (length)
from the root to a leaf node is N. In other words, this means a member or user might need to perform
N-1 decryptions before the valid user gets the group key and store log k (N). Some members may need
to hold the number of keys in an unbalanced key tree, N keys, while others might need to store just two
keys. Individual rekeying, i.e., rekeying for any request to join or leave, has two disadvantages. First, it is
inefficient in terms of computation and communication overhead as rekeying messages have to perform
authentication; a high rate of join/departure requests will lead to the degradation of performance as the
signing process may sometimes get computationally costly. Second, suppose there is a high delay in
rekeying message transmission and regular join /departure request rates. In that case, a participant can
need a large amount of memory space to store rekeying and data messages that can not decrypt. Batch
rekeying has been proposed to mitigate these problems by commercializing between performance and
safety. Under this scheme, participants join and leave over a period of time before rekeying is conducted
are consolidated. We intend to apply the algorithms that are appropriate for a batch join event within
the proposed work. The algorithms balance the main tree and reduce its rekeying costs, which lead
to savings in bandwidth relative to the existing algorithms. Having a balanced key tree for each batch
greatly benefits mobile devices since they usually have restricted storage and processing power. The ever-
increasing number of Internet applications, such as software and content distribution, pay-per-view, video
conferencing, real-time information services, requires effective and reliable communication platforms
with multicast. There are many security technologies designed for unicast communication,they are not
feasible multicast behaviour. Alternatively, frameworks are adopted which use a secure model of group
communication. Under such a model, the multicast group members share an asymmetric key, here called
the session key, used to encrypt traffic in the multicast group. If group membership changes during the
departure of any valid user, the group needs to be rekeyed accordingly, and the new members should not
be given access to previous sessions or group keys (backward secrecy). The members left should not
be given access to a future session or group keys (forward secrecy). The rekeying caused by leaving the
members can be very costly linear to the group size in static key management. Therefore, scalable key
management is a significant problem that needs to be considered to facilitate safe multicasting for large
114
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
and diverse groups. Yet reliability, achieving scalability, and pulling the security specifications in various
directions, and balancing them to suit the application needs is still an open problem. Several approaches
were proposed for the scalable multicast key management schemes based on the Logical Key Hierarchy
(LKH) and its variants. The schemes are effective in rekeying costs and key storage requirements for each
group member. The storage requirements are still high for the group controller and individual members,
rising linearly with the number of users. Individual rekeying, i.e., rekeying after every request to enter
or leave, has two disadvantages. The problem of delay and requirement of signed authentication leads
to batch rekeying with hybrid rekeying in Logical Key Hierarchy (LKH). Logical key hierarchy and its
variants are analyzed with a centralized or distributed approach; each method’s work is presented along
with its properties. Compared with the unicast-based approach, the hierarchical key distribution ap-
proach, LKH, is scalable and efficient group rekeying compared with the unicast-based approach. OFT
is effective on bandwidth compared to LKH. OFC, IHC, and SDLKH reduce overhead communication
than LKH, mainly in the operation of leave users. Distributed LKH and LKH Diffie-Hellman every the
number of keys each user holds. We classify them as Pair-wise schemes and Group-wise schemes based
on whether the key is generated for pair-wise nodes or group members. The key management is classi-
fied as centralized and distributed based on whether the generation, maitainence of the cryptographic
group key have been allocated to a single node or multiple nodes. In centralized schemes, a single entity
generates the group key. That entity also performs access control and key distribution.
The main drawback is that the central server becomes a single point of failure. If the key server fails,
the complete group will be affected. The group becomes vulnerable if key servers get compromised. Also,
if a group becomes large, then scalability will also become an issue. The performance of the centralized
system depends on storage requirements, size of messages. The size of messages is characterized by the
number of bytes in a rekey process for the join and leave operation. Secondly, backward and forward
secrecy is also to be maintained. Collusion should also be maintained. Expelled members should not
be able to regenerate the group key by sharing keying information, among others. Based on whether
node keys are updated or not during the whole lifetime, we classify the schemes as static and dynamic
schemes. Based on whether location or deployment knowledge is considered a parameter during the
key pre-distribution process, we classify the Location-dependent and Location-independent schemes.
Authors classified key management schemes for WSNs mainly based on the storage overhead per sensor
node (Dave, 2016). They are classified into three categories based on the storage incompetent, storage
competent, and highly storage competent key management schemes. Some of the schemes have focused
on diminishing storage space requirements. The authors give a probabilistic key distribution scheme
that launches pair wise keys between neighbouring sensor nodes. The random key pre-distribution rule
(RKP), it is mainly classified into three-level known as key pre-distribution, discovering shared key, and
path key formation. In the key pre-distribution phase, each sensor node randomly chose m keys from a
key pool S. Each sensor node discovers the keys shared with its neighbours in the shared key discovery
phase (Eschenauer and Gligor, 2002). If two neighbouring sensor nodes have no shared key identifiers
during the shared key discovery process, they will form a shared key via two or more hops during the
path key formation phase. The authors proposed a secured broadcast/multicast scheme for ranked sensor
networks. The scheme can guard against the common node capture attacks of wireless networks. Altering
the session key every time can also attain flawless forward secrecy in this scheme (Chen, Huang, Lin
2012). Lightweight key management based on initial trust is given in (Dutertre Cheung, & Levy 2004).
A survey-based on dynamic key management is given in (Eltoweissy, Moharrum & Mukkamala 2006).
Location-based key management-based pairwise keys are given by (Liu & Ning 2003).
115
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
Key management based on deployment knowledge is given by (Du, Deng, Han, Chen & Varshney
2004). In group key management, a trusted server generates the group key whenever a new member
enters into a system. It will send that key to new members using the unicast method and other members
using multicast manner using group key. Forward and backward secrecy is the most important aspect to
be maintained in group key management. thus, changing the group key requires changing and informing
all members. If a new member joins, he should not receive old messages; thus, to maintain backward
secrecy, the group key is changed. When a member leaves, he should not be able to decrypt old messages;
thus, forward secrecy has to be maintained. The old group key is to be cannot be used in a multicast
manner to obtain a new key during leave. Thus, leave operation is more complex than the join process.
It leads to more communication and computation overhead. Thus, if a system is dynamic, scalability
becomes a very important issue in group key management. Tree-based key management is given in
(Messai, Aliouat,& Seba 2010).
This chapter focuses on a comprehensive overview of different group key management schemes along
with key management solutions. This chapter is organized as follows: We give an overview of different
group key management schemes. Section 2 presents various schemes for secure multicast. In section
3, we present our proposed scheme based on a simple logical key hierarchy (LKH). Section 4 gives a
result analysis of the proposed scheme. Section 5 presents the conclusion along with the future scope.
BACKGROUND
In group key communication, communication within a group is visible and accessible to its group
members only. It is done by generating a cryptographic key generated by the group manager, which
is distributed to all valid members of the group. Whenever a node joins the group, the node is given a
group key. It is also ensured that it should be able to decipher the backward messages among the group.
This is backward secrecy. When a node leaves the group, the also group key has to be renewed. It will
prevent the new node from getting future messages.
This is called forward secrecy. This requires that group key management should be scalable, dy-
namic, and efficient. In paper (Duma et al., 2003), secure multicast communication is presented for
pay per view and secure video conferencing. A tree-based approach is used to derive a group key that
cost scales with the group size logarithm. In the proposed scheme, two merging algorithms are used.
To increase efficiency, batch rekeying is used. This results in a balanced key with reduced rekeying
116
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
cost. In paper (Pande & Thool, 2016), efficient group management with efficient rekeying is given.
The GREP algorithm is presented that leverages the history of joins to make the scheme more efficient
and scalable. The overhead is also independent of group size. It also has the advantage it easily recov-
ers from collusion attacks. It is implemented in Contiki OS. It is shown that GREP is efficient, good
scalability, and deployable on the resource-constrained nodes (Tiloca, Dini, Rizki, Raza 2019). A group
key management scheme for encrypted multicast traffic is transmitted for satellite communication. In
the proposed scheme, a Logical key hierarchy is used for member registration and periodic admission
(Howarath, Iyengar, Sun, Cruickshank 2004). It results in reduced cost and optimum out-degree. It also
increases network utilization. But In a real-scenario, we have a large multicast group where node leaves
and joins occur frequently, and also the number of nodes leaving and joining is also not proportionate.
Hence Scalable rekeying is an important issue that needs to be concerned for the Secured Group
communication for dynamic groups.. Applications such as teleconferences, information services, distrib-
uted interactive simulations use group communications. Hence, providing confidentiality, authenticity,
and integrity of messages becomes an important factor in determining the group model’s performance.
A novel solution to scalability is provided where the secure group is presented as a triple (U, K, R)
where U is a set of users, K is the set of keys, and R stands for user key relationships. Key graphs are
used to denote secure groups. Three strategies are presented to secure distributed rekey messages (Wong,
Gouda, Lam 2000).
In secure group communication, members share a unique cryptographic key. Rekeying process
is frequently invoked in dynamic groups that require an efficient group key generation. A group key
agreement is proposed that exploits the state vectors of group members. State vectors is defined as a
set of randomly generated nonces representing a logical link between group members. It enables group
members to generate multiple cryptographic keys independently. Thus, with local knowledge of a secret
once, each member can generate and share a number of secure keys, which makes it more secure and
efficient in computation and communication (Mohammad & Shin 2017).
A logical key hierarchy is used to generate the group key in multicast group rekeying. However, it
does not consider the changing behavior or probabilities of leave and joining different group members.
It makes a basic scheme unsuitable for dynamic groups. Based on the basic behaviour of different users,
the active members and inactive members are partitioned. They are set on different locations on a key
hierarchy. it leads to a decrease in the number of encryptions in the group manager and ultimately leads
to reduced communication overhead (Yong & Sun 2005).
A survey on LKH is presented with performance evaluation for LKH and its variant key management
methods. Among LKH and its variant key management methods in (Wong et al., 2000), hierarchical
methods are presented, which work with a centralized or distributed approach and are close to LKH
results. Each form of Hierarchical Key Distribution works differently (Ng, W. H. D., Howarth, M., Sun,
Z., & Cruickshank 2007). The difference is studied and discussed in maintaining the Hierarchical Key
Tree. LKH and its variants are studied to know each method’s functioning, its core purpose, computa-
tion, and communication method, and its differences from each other, know which method is appropriate
for an application, and summarize the comparison of these variants with LKH. The author proposes a
two-level Hybrid Key Tree multicast key management scheme. The entire multicast group of users is
partitioned into clusters of similar size. Global level controls cluster membership in the multicast com-
munity, and cluster-level controls user membership in clusters. The various mapping of encrypted group
keys to nodes in the key tree and the different key management algorithms used for global and cluster
117
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
level controlling allow the two levels in HKT to have efficient security and performance characteristics
(Ng, W. H. D., Cruickshank, H., & Sun, Z. 2006).
Group key
Server cost Join Leave 1-affects n problem
management scheme
LKH O(logN) d+1 2d yes
OFT O(logN)/2 d+1 d+1 yes
OFC O(logN) logdN +1 (d − 1)logdN yes
IHC/ SD-LKH O(logN) 1+logdN (d − 1)logdN yes
Distributed LKH log2Nd - - yes
Diffie–Hellman LKH log2N+1 - - yes
REVIEW OF LITERATURE
A centralized approach that is Logical key hierarchy is a key management and distribution scheme where
a key is obtained at three different levels: individual, intermediate, and group level. Probably, it can ap-
ply to all hierarchical tree structures. The number of messages expected for the process of rekeying is
decreased related to GKMP(Pande & Thool, 2016)
In OFT, for the key generation, the bottom-up approach is used. It is a tree-based structure to overcome
rekeying costs. For every single operation, each time key is determined at the leave node and broadcast
to the root, minimizing the storage and computation cost. This algorithm key is communicated to the
root level(Sherman & McGrew, 2003)
NSGC shares keys within nodes and provides security with very little cost. Each node’s own individual
key or unique key is used to calculate the group communication’s static group key.(Kumar & Lavanya,
2015)KMSGC is an improved protocol based on a grouping system inside the cluster moreover gives an
explanation for various problems, e.g., reduced scalability, delivery time, and low operating frequency.
(Bao et al., 2014)
118
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
To ensure effective delivery of group key within the structure, an advanced protocol is used that de-
pends on an enhanced key distribution algorithm. Limiting the time and storage requirements statically
generates variable-length UID using one-way hash function (Kumar et al., 2013).
For ensures key communication during rekeying, a probabilistic security-based scheme is used be-
tween any set of two nodes or nodes with base station taking the constraints of cost (Albakri et al., 2019).
Dynamically symmetrical and balanced tree approaches are now used to save and manipulate the
keys. The balancing of the tree remains the same in any member who joins/leaves in the group. Two
merging algorithms are used for the design of this protocol: merging algorithms and batch balanced
Algorithm (Ng et al., 2007).
The ternary tree-based protocol always handles the issues that appear in secure multicast communica-
tion and highlights the adoption of reasonable procedures to solve those issues (Vijayakumar et al., 2013).
PROPOSED WORK
In this section, we proposed a new algorithm for rekeying. Logical Key Hierarchy works quite optimal
for small group communication. But when considering Large scalable groups, Logical Key Hierarchy
performance degrades as the number of members increases beyond a particular height. Taking advantage
of the Logical Key Hierarchy efficiency over small groups, we have consolidated the Logical Key Hier-
archy and the Linked List concept. We maintained a Logical Key Hierarchy up to an optimum height and
kept the pointer to the Logical Key Hierarchy’s root in a Linked List. Let say ‘k’ be the optimum height
of the Logical Key Hierarchy for which LKH rekeying based algorithms works in an optimum manner.
Now for the height ‘k’ we can have the maximum number of members in a group = 2k. Hence, a new
LinkedIn node is created for the rest members, and a new LKH is constructed. Forward and Backward
secrecy is maintained in the same way as in the normal LKH. It can be seen that the number of group
members is significantly less affected as compared to normal LKH as we have restricted the height
of the tree. The algorithm works in three phases Initialization, Joining, and Leaving the Group. In the
Initialization phase, we construct the LKH tree based on the static number of Group members. Joining
and Leaving phase are used for the Dynamic working of Linked LKH. In the Joining phase, the New
joining members request the GC, and in the Leaving Phase, the Group members who wished to leave
the Group are processed. It can be noted that If the Leave and Join phase can occur simultaneously, then
the Leave process is given priority first to work.
Symbol Meaning
N Total number of group members.
J Total number of joining members.
D Total number of departing members.
K Optimal Height of LKH tree
119
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
Initialization
Joining
1. Scan the Linked List and compare the space available in the corresponding LKH.
2. If space is found then we simply insert the joining members.
3. Recursively update the keys from new Joined node members up to root using the keygenerator.
4. Balance the LKH tree if necessary.
5. Update the Member_Available Hash-set with the new joining members.
6. If no space is found during Step 1) then, we extend the Linked List and insert the new joining
members as in the Initialization phase.
Leaving
1. The Members who are to leave, sent the leave request to the Group Controller.
2. Find the Linked List node by using the Member_Identify Hash-map (called by GC).
3. Remove the member node from the corresponding LKH tree.
4. Update the MemberId and key used by the removed node.
5. Backtrack the path from the removed node to the root and update the keys using the KeyGenerator
(called by GC)
6. Update the Member_available Hash set for the Linked list node.
7. Repeat the above procedure until all the request removed nodes are processed.
120
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
Figure 2. Represents a linked logical key hierarchical for 12 group members represented by U1 –U12.
Figure 3. Represents the data flow diagram (DFD) for linked logical key hierarchy algorithm
RESULT ANALYSIS
A Logical Key Hierarchy is proved to be a Memory Efficient approach than Group Rekeying Protocol.
However, the Normal LKH approach still suffers an overhead of rekeying the LKH tree for a large group.
We can find that Logical Key Hierarchy rekeying depends on the tree’s height and thus indirectly over
the number of Group members. Effective rekeying cost for a Normal LKH = log(N) where ‘N’ is the
Group members’ number. Our Proposed Linked LKH algorithm performs rekeying efficiently than the
normal LKH algorithm, taking the Optimal Height ‘K’ and making this height as a constraint over the
maximum height results in less rekeying update messages traversal within the tree. Considering the
Hash-set working complexity as Constant ‘C’ for finding the LKH corresponding to a new joined or
leaving the node.
121
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
As shown in Figure 4, the graph represents the relation between the number of nodes in the tree (x-
axis) and the effective rekeying cost (y-axis) for the Normal LKH and Linked LKH algorithm.
It is observed from the graph, as shown in Fig. 6, that Linked LKH rekeying cost is much lesser than
the traditional LKH. However, Fig. 4 describes our proposed algorithm’s linear complexity, as we need
to traverse the Hashmap to find the Empty available space. However, we benefit from limiting the height
of LKH when a node is going to depart. Fig. 6 depicts the same. Since each LKH inside the Linked List
node is similar to the normal LKH, our algorithm also ensures forward and backward secrecy.
Figure 4. Graph between effective rekeying cost vs. number of group members
122
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
CONCLUSION
This paper has proposed a new LinkedIn LKH algorithm that is more efficient than the existing traditional
LKH algorithm. Since the proposed Linked LKH algorithm is based on the LKH, it has less storage
overhead than the GREP. The effective rekeying cost of the Linked LKH algorithm is Constant, which
is better than the traditional LKH, which depends on the number of Group members’ logarithmic factor.
We have also graphically observed our algorithm and the LKH algorithm by considering the effective
rekeying cost. However, there is still Initialization complexity overhead for the Linked LKH, which
increments directly with Group members. The Initialization complexity is not so important if we have
already constructed a tree by using the method like Pre-registration of the Group Members. As a Future
work, the Initialization phase complexity can be made to be concerned, which will lead to an overall
less effective cost for constructing and maintaining the LKH.
REFERENCES
Albakri, A., Harn, L., & Song, S. (2019). Hierarchical key manage- ment scheme with probabilistic
security in a wireless sensor network (WSN). Security and Communication Networks, 2019.
Bao, X., Liu, J., She, L., & Zhang, S. (2014, June). A key management scheme based on grouping
within cluster. In Proceeding of the 11th World Congress on Intelligent Control and Automation (pp.
3455- 3460). IEEE.
Barskar, R., & Chawla, M. (2016). A survey on efficient group key management schemes in wireless
networks. Indian Journal of Science and Technology, 9(14), 1–16. doi:10.17485/ijst/2016/v9i14/87972
Bilal, M., & Kang, S. G. (2017). A secure key agreement protocol for dynamic group. Cluster Comput-
ing, 20(3), 2779–2792. doi:10.100710586-017-0853-0
123
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
Chen, C. T., Huang, S. Y., & Lin, I. C. (2012). Providing perfect forward secrecy for location-aware
wireless sensor networks. EURASIP Journal on Wireless Communications and Networking, 2012(1),
241. doi:10.1186/1687-1499-2012-241
Chen, C. Y., & Chao, H. C. (2014). A survey of key distribution in wireless sensor networks. Security
and Communication Networks, 7(12), 2495–2508. doi:10.1002ec.354
Dave, M. (2016, March). Storage as a parameter for classifying dynamic key management schemes
proposed for WSNs. In 2016 International Conference on Computational Techniques in Information
and Communication Technologies (ICCTICT) (pp. 51-56). IEEE.
Du, W., Deng, J., Han, Y. S., Chen, S., & Varshney, P. K. (2004, March). A key management scheme
for wireless sensor networks using deployment knowledge. In IEEE INFOCOM 2004 (Vol. 1). IEEE.
Duma, C., Shahmehri, N., & Lambrix, P. (2003, June). A hybrid key tree scheme for multicast to balance
security and efficiency requirements. In WET ICE 2003. Proceedings. Twelfth IEEE International Work-
shops on Enabling Technologies: Infrastructure for Collaborative Enterprises, 2003 (pp. 208-213). IEEE.
Dutertre, B., Cheung, S., & Levy, J. (2004). Lightweight key management in wireless sensor networks
by leveraging initial trust. Technical Report SRI-SDL-04-02, SRI International.
Eltoweissy, M., Heydari, M. H., Morales, L., & Sudborough, I. H. (2004). Combinatorial optimization
of group key management. Journal of Network and Systems Management, 12(1), 33–50. doi:10.1023/
B:JONS.0000015697.38671.ec
Eltoweissy, M., Moharrum, M., & Mukkamala, R. (2006). Dynamic key management in sensor networks.
IEEE Communications Magazine, 44(4), 122–130. doi:10.1109/MCOM.2006.1632659
Eschenauer, L., & Gligor, V. D. (2002). A key-management scheme for distributed sensor net-
works. Proceedings of the 9th ACM Conference on Computer and Communication Security, 41–47.
10.1145/586110.586117
He, X., Niedermeier, M., & De Meer, H. (2013). Dynamic key management in wireless sensor networks:
A survey. Journal of Network and Computer Applications, 36(2), 611–622. doi:10.1016/j.jnca.2012.12.010
Howarth, M. P., Iyengar, S., Sun, Z., & Cruickshank, H. (2004). Dynamics of key management in secure
satellite multicast. IEEE Journal on Selected Areas in Communications, 22(2), 308–319. doi:10.1109/
JSAC.2003.819978
Jain, A. K., & Gupta, B. B. (2019). A machine learning based approach for phishing detection using
hyperlinks information. Journal of Ambient Intelligence and Humanized Computing, 10(5), 2015–2028.
doi:10.100712652-018-0798-z
Kumar, N. S., & Lavanya, S. (2015). A novel scheme for secure group communication in multicast
network. International Journal of Security and Networks, 10(2), 65–75. doi:10.1504/IJSN.2015.070409
Kumar, S., Purusothaman, T. N. M., & Lavanya, S. (2013). Design and performance analysis of scal-
able and efficient group key Manage- ment scheme [SEGKMS] for group communication in multicast
net- works. Life Science Journal, 10(2).
124
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
Liu, D., & Ning, P. (2003, October). Location-based pairwise key establishments for static sensor net-
works. In Proceedings of the 1st ACM workshop on Security of ad hoc and sensor networks (pp. 72-82).
10.1145/986858.986869
Messai, M. L., Aliouat, M., & Seba, H. (2010). Tree based scheme for key Management in wireless
sensor networks. EURASIP Journal on Wireless Communications and Networking, 2010(1), 1–10.
doi:10.1155/2010/910695
Ng, W. H. D., Cruickshank, H., & Sun, Z. ((2006) Scalable balanced batch rekeying for secure group
communication. Computers & Security, 25(4), 265-273.
Ng, W. H. D., Howarth, M., Sun, Z., & Cruickshank, H. (2007). Dynamic balanced key tree management
for secure multicast communications. IEEE Transactions on Computers, 56(5), 590–605. doi:10.1109/
TC.2007.1022
Ouaguid, A., Abghour, N., & Ouzzif, M. (2018). A novel security framework for managing android
permissions using blockchain technology. International Journal of Cloud Applications and Computing,
8(1), 55–79. doi:10.4018/IJCAC.2018010103
Pande, A. S., & Thool, R. C. (2016, September). Survey on logical key hierarchy for secure group com-
munication. 2016 International.
Rafaeli, S., & Hutchison, D. (2003). A survey of key management for secure group communication.
ACM Computing Surveys, 35(3), 309–329. doi:10.1145/937503.937506
Rahman, M., & Sampalli, S. (2015). An efficient pairwise and group key management protocol for wireless
sensor network. Wireless Personal Communications, 84(3), 2035–2053. doi:10.100711277-015-2546-4
Seetha, R., & Saravanan, R. (2015). A survey on group key management schemes. Cybernetics and
Information Technologies, 15(3), 3–25. doi:10.1515/cait-2015-0038
Sherman, A. T., & McGrew, D. A. (2003). Key establishment in large dynamic groups using one-way func-
tion trees. IEEE Transactions on Software Engineering, 29(5), 444–458. doi:10.1109/TSE.2003.1199073
Stergiou, C., Psannis, K. E., Gupta, B. B., & Ishibashi, Y. (2018). Security, privacy & efficiency of
sustainable cloud computing for big data & IoT. Sustainable Computing: Informatics and Systems, 19,
174–184. doi:10.1016/j.suscom.2018.06.003
Tiloca, M., Dini, G., Rizki, K., & Raza, S. (2019). Group rekeying based on member join history. Inter-
national Journal of Information Security, 1–39. doi:10.100710207-019-00451-0
Vijayakumar, P., Bose, S., Kannan, A., & Jegatha Deborah, L. (2013). Computation and Communication
Efficient Key Distribution. Academic Press.
Wallner, D., Harder, E., & Agee, R. (1999). Key management for multicast: Issues and architectures.
RFC 2627.
Wong, C. K., Gouda, M., & Lam, S. S. (2000). Secure group communications using key graphs. IEEE/
ACM Transactions on Networking, 8(1), 16–30. doi:10.1109/90.836475
125
Scalable Rekeying Using Linked LKH Algorithm for Secure Multicast Communication
Xu, Y., & Sun, Y. (2005, December). A new group rekeying method in secure multicast. In International
Conference on Computational and Information Science (pp. 155-160). Springer. 10.1007/11596981_23
Yousefpoor, M. S., & Barati, H. (2019). Dynamic key management algorithms in wireless sensor net-
works: A survey. Computer Communications, 134, 52–69. doi:10.1016/j.comcom.2018.11.005
Zheng, Q., Wang, X., Khan, M. K., Zhang, W., Gupta, B. B., & Guo, W. (2017). A lightweight authen-
ticated encryption scheme based on chaotic scml for railway cloud service. IEEE Access: Practical
Innovations, Open Solutions, 6, 711–722. doi:10.1109/ACCESS.2017.2775038
Zhu, S., & Jajodia, S. (2003, December). Scalable group rekeying for secure multicast: A survey. In
International Workshop on Distributed Computing (pp. 1-10). Springer. 10.1007/978-3-540-24604-6_1
126
127
Chapter 8
Botnet Defense System and
White-Hat Worm Launch
Strategy in IoT Network
Shingo Yamaguchi
Yamaguchi University, Japan
Brij Gupta
National Institute of Technology, Kurukshetra, India
ABSTRACT
This chapter introduces a new kind of cybersecurity system named botnet defense system (BDS) that
defends an IoT system against malicious botnets. This chapter consists of two parts. The former part
describes the concept and design of the BDS. The concept is “fight fire with fire.” To realize the concept,
the BDS uses bot technology. The BDS builds a white-hat botnet on the IoT system by itself and uses it
to exterminate the malicious botnets. The white-hat botnet autonomously spreads over the IoT system
and thus drastically increases the defense ability. The latter part explains the strategy of the BDS. The
white-hat botnet is a so-called double-edged sword. It defends the IoT system against malicious botnet
but wastes the system’s resources. Therefore, the BDS should strategically use the white-hat botnet.
Some strategies have been proposed. Their characteristics are discussed through the simulation with
the agent-oriented petri nets.
INTRODUCTION
Internet of Things (IoT) aims everything including humans to interact and to create new values from shar-
ing information. IoT has enriched our lives while gives rise to a new risk on cybersecurity. IoT devices
are explosively increasing, and the number is predicted to reach 30 billion by 2023 (Cisco, 2020). The
problem is that most of them are vulnerable. This is because they do not have resources to run security
functions and their vendors may sacrifice security in the price competition and/or their rush to market.
In September 2016, that risk became reality. IoT was used as a springboard of giant distributed denial-
DOI: 10.4018/978-1-7998-7789-9.ch008
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
of-service (DDoS) attacks, which struck many of the world’s biggest sites such as Netflix and Twitter
(O’Brien, S.A.,2016). These attacks were brought about by malware called Mirai. Mirai infects IoT
devices and turns them into bots. Those bots form a network (botnet) that can be used for DDoS attacks.
For the detail of Mirai, refer to (Sinaović, H., & Mrdovic, S., 2017) and (Yamaguchi, S. & Gupta, B.,
2019). Mirai’s DDoS attacks have a tendency to be large-scale and disruptive. This is because IoT devices
are characterized by large-volume, pervasiveness, and high vulnerability (Kolias, C., Kambourakis, G.,
Stavrou, A., & Voas, J., 2017). Mirai has spread to emerging markets and developing countries (Nakao,
K., 2018). In early October 2016, Mirai infected over 300,000 IoT devices in 164 countries (Devry, J.,
2016). To make matters worse, Mirai’s authors published the source code (Bonderud, D., 2016). It gave
rise to many variants of Mirai such as Satori (360 netlab., 2017) and Okiru (Arzamendi, P., Bing, M. &
Soluk, K., 2018). Even now after five years since Mirai appeared, Mirai and variants continue to rage
all over the world (Milić, J., 2019).
Some techniques have been proposed against Mirai’s threat. The United States Computer Emergency
Readiness Team (US-CERT) showed rebooting the infected device can clear Mirai (US-CERT, 2016).
This is because Mirai penetrates only to the dynamic memory of the device. However, Moffitt, T. (2016)
reported that Mirai can reinfect the device within minutes unless the vulnerability is patched. The other
techniques can be roughly divided into three categories: detection, mitigation, and spread prevention.
The following are typical examples.
Detection Techniques: Bezerra, V.H., da Costa, V.G.T., Barbon, J., Miani, R.S., & Zarpelão, B.B.
(2019) have proposed a host-based approach to detect IoT botnets called IoTDS (Internet of Things De-
tection System). IoTDS monitors a device and collects its CPU use and temperature, memory consump-
tion, and the number of processes. If the device detects any anomaly from the data, an alert of botnet
detection is sent to the central server.
Meidan, Y., Bohadana, M., Mathov, Y., Mirsky, Y., Shabtai, A., Breitenbacher, D., & Elovici, Y.
(2018) have proposed a network-based anomaly detection method for the IoT called N-BaIoT. N-BaIoT
extracts behavior snapshots of the network and uses deep autoencoders to detect anomalous network
traffic from compromised IoT devices.
Mitigation Techniques: Some of this category include both detection and mitigation (Jaramillo,
L.E.S., 2018) and (Alomari, E., Manickam, S., Gupta, B. B., Anbar, M., Saad, R. M., & Alsaleem, S.,
2016). Manso, P., Moura, J., & Serrão, C. (2019) have proposed a Software-Defined Intrusion Detection
System. This system can automatically detect several DDoS attacks. Once the IDS detects an attack,
it notifies a software-defined networking controller to control devices. Therefore, it timely enables to
detect a botnet exploitation, to mitigate malicious network traffic, and to protect normal network traffic.
Ceron, J.M., Steding-Jessen, K., Hoepers, C., Granville, L.Z., & Margi, C.B. (2019) have proposed
a network layer that adapts itself to mitigate the network traffic generated by malware. It can modify the
traffic at the network layer based on the actions performed by the malware.
Spread Prevention Techniques: Gopal, T.S., Meerolla, M., Jyostna, G., Eswari, L., Reddy, P., &
Magesh, E. (2018) have proposed a whitelisting based solution to prevent Mirai from spreading. They
showed the successful blocking of Mirai malware through the experiment.
Frank, C., Nance, C., Jarocki, S., & Pauli, W. E. (2018) have proposed two scripts executable on
actual devices to protect devices from becoming Mirai bots. They show that the hardening script was
shown to be successful in preventing the initial Mirai infection on the device and the detection script was
successful in recognizing and stopping an already existing infection in a controlled test environment.
128
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
These techniques are useful to detect Mirai bots and to mitigate their threat. However, they are not
radically something to exterminate Mirai bots. As described before, the simplest solution is rebooting
the infected device and patching its vulnerability. However, patching by manpower tactics is unrealistic
because IoT devices are explosively increasing. Thus, an innovative approach is needed to drastically
increase the defense ability against Mirai.
This chapter introduces a new kind of cybersecurity system named Botnet Defense System (BDS)
that exterminates malicious botnets (Yamaguchi, S., 2020a). After the introduction in Section 1, Section
2 describes the concept and design of BDS. The BDS adopts the concept of “fight fire with fire” and
makes use of botnet technology for defense. The BDS builds a white-hat botnet on the IoT system by
itself and uses it to exterminate the malicious botnets. The white-hat botnet autonomously spreads over
the IoT system and thus drastically increases the defense ability. Section 3 explains the strategy of the
BDS. The white-hat botnet is a so-called double-edge sword. It defends the IoT system against mali-
cious botnet but wastes the system’s resources. Therefore, the BDS should strategically use the white-hat
botnet. Some strategies have been proposed. Their characteristics are discussed through the simulation
with the agent-oriented Petri nets. Section 5 summarizes key points and gives future research directions.
This section describes the concept and design of the BDS and the white-hat worm used in the BDS.
Concept
Botnet technology (Bailey, M., Cooke, E., Jahanian, F., Xu, Y., Karir, M., 2009) has been widely used to
conduct malicious activities like DDoS attacks. Ironically, since Mirai’s source leaked, this technology
has been more accelerating evolution and has produced various variants. The purpose of the variants is
129
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
no longer limited to DDoS attacks but extends to cryptocurrency mining (Arghire, I., 2017) and proxy
function (Cimpanu, C., 2018). In October 2016, a new type of worm called Hajime was found (Edwards,
S.,& Profetis, I, 2016). Like Mirai, Hajime infects IoT devices and turns them into bots. Hajime bots
block the ports that Mirai uses to infect and prevent Mirai from spreading. These prior cases suggest
a new possibility of the botnet technology. That is, the botnet technology could become an innovative
solution to defend an IoT system against malicious botnets.
The concept of the BDS is “fight fire with fire”. This means to use the same weapon that attackers
use. The BDS realizes this concept and fights malicious botnets with a botnet. Concretely, it builds a
botnet on the IoT system by itself and uses it to exterminate the malicious botnets. The botnet built by
the BDS is used for well-intentioned purposes and thus is called a white-hat botnet. The white-hat botnet
autonomously spreads over the IoT system and thus drastically increases the defense ability.
Design
The BDS adopts component-based architecture. This enables us to research and develop the function-
alities using components as a unit and further to realize a required instance quickly and flexibly by
combining components.
Figure 1 shows the system configuration. The left rounded rectangle represents the BDS. The BDS
consists of four components: monitor, strategy planner, worm launcher, and command and control
(C&C) server.
• Monitor component: watches over the IoT system. This activity itself may be done through white-
hat bots. If detecting a malicious botnet, it investigates and reports the information such as the
botnet type and its infection situation.
• Strategy planner component: plans a strategy against the malicious botnet based on the informa-
tion reported by the monitor component.
• Worm launcher component: sends white-hat worms into the IoT system based on the strategy and
builds a white-hat botnet.
• C&C server component: controls the white-hat botnet to exterminate the malicious botnet.
The right rounded rectangle of Figure 1 represents the IoT system defended by the BDS. Its inside
graph represents the network of the IoT system. Bot malware searches for the next victim from the entire
Internet. We project the relationship between victims on the IoT system and consider it as the network of
the system. This network consists of 12 nodes and has full mesh topology. The concept of observability
and controllability is introduced to the network (Yamaguchi, S., 2021). For simplicity, assume in this
chapter that every node is observable and controllable. That is, the BDS can check whether each node
has a bot, and can send a white-hat worm to any node without a malicious bot.
Let us illustrate an operation of the BDS with the example of Figure 1.
1. Monitoring: The BDS checks whether each node has a bot. As a result, the BDS found out five
malicious bots in total at the 1st, 2nd, 3rd, 4th, and 5th nodes. The black circles (●) of Figure 1
represent malicious bots.
2. Strategy Planning: The BDS plans a strategy to launch white-hat worms. Based on the IoT system’s
status, the BDS adopted the Few-Elite strategy described later.
130
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
3. Worm Launching: Based on the strategy, the BDS sends white-hat worms and builds a white-hat
botnet. It sent three white-hat worms in total to the 8th, 9th, and 10th nodes. The white circles (
〇) of Figure 1 represent white-hat bots produced by those worms.
4. C&C: The BDS controls the white-hat botnet to exterminates the malicious bots. The BDS will
finally exterminate the malicious botnet.
The distinctive feature of the BDS is to use white-hat worms and their botnets to fight against malicious
botnets.
Some worms are not used for malicious activities, but attempt to protect the device from other malware
instead. Such worms are called white-hat. An early example of white-hat worms is Linux.Wifatch. It
infects a device and disables Telnet to keep other malware out (Ballano, M., 2015). Another representa-
tive example is Hajime. It turns devices into bots but possesses no capability for DDoS attacks. It blocks
the ports that Mirai uses to infect and protects them against Mirai. However, because not systematically
controlled, Hajime botnet continues to spread even after completing the defense against Mirai.
Molesky, M.J., & Cameron, E.A. (2019) have proposed not only the manufacturers but also govern-
ments to use white-hat worms to fix critical vulnerabilities of IoT devices. They state that this could be
enacted by including explicit terms within the Terms and Conditions agreement at the time of purchase
and creates a contract with the consumer allowing for these actions to occur legally and without liability
to the company. This suggests white-hat worms are applicable in practice.
The white-hat worm used in the BDS aims to build a time-limited white-hat botnet to exterminate
malicious botnets (Yamaguchi, S., 2020b). The worm’s capabilities are represented by two attributes.
• Secondary infection possibility: The worm infects even the device infected by bot malware, i.e. it
infects the botized device, and then removes the malicious bot from the device. Secondary infec-
tion possibility is the probability that the worm infects the botized device.
• Lifespan: The worm has a lifespan, i.e. it removes itself when exhausting the lifespan. Even after
removing the worm, the infected device still is a white-hat bot. It exhibits an immune effect against
malware until reboot.
A Mirai variant called OMG possesses proxy functions (ASERT Team., 2018). The white-hat bots
used in the BDS also incorporate the proxy functions. They enable the BDS to indirectly access unob-
servable and uncontrollable nodes, e.g. inside private networks (Yamaguchi, S., 2021).
STRATEGY
This section describes the strategies used in the BDS. Once the BDS detects a malicious botnet, it plans
a strategy for exterminating the botnet. A good strategy results in successful extermination, vice versa.
Therefore, strategy studies are essential to produce intended results. Strategies can be roughly divided
into two categories.
• Worm launch strategies: specify how to send white-hat worms to build its botnet.
131
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
• C&C strategies: specify how to command and control the botnet to exterminate the malicious
botnet.
This chapter focuses on the worm launch strategies. For those who are interested in the C&C strate-
gies, refer to Reference (Yamaguchi, S., 2021).
A worm launch strategy L is formalized as a mapping from a monitoring information (RMirai, 𝛿, (#nodes,
Ntopology, Ndensity), (ℓ, 𝜌)) to the number #White of the white-hat worms to launch, i.e.
: RMirai , , #nodes , N topology , N density , , #White (1)
where
• Malware is Mirai.
• RMirai is a Mirai’s infection rate and is given as RMirai= #Mirai/#nodes, where #Mirai is the number of Mirai
bots, #nodes is the number of network nodes, and each node has one device.
• 𝛿 is the delay time until an infected device is rebooted.
• (#nodes, Ntopology, Ndensity) is the IoT system’s specification. Ntopology is the network topology such as
mesh and tree. Ndensity is the network density and is given as
2 # AC
N density (2)
#nodes #nodes 1
• (ℓ, 𝜌) is the white-hat worm’s capability, where ℓ is the lifespan and 𝜌 is the secondary infection
possibility.
All-Out Strategy
The All-Out strategy All −Out is to launch as many white-hat worms as possible. It is defined as a map-
ping such that
When adopting this strategy, the BDS sends white-hat worms to all nodes that do not have Mirai bots
regardless of the information (RMirai, 𝛿, (#nodes, Ntopology, Ndensity), (ℓ, 𝜌)).
Figure 2 shows an application example of the All-Out strategy to the IoT system of Figure 1. This
system has a network composed of 12 nodes, i.e. #nodes=12. Figure 2a shows the state when the BDS
detected the Mirai botnet composed of five Mirai bots, i.e. #Mirai=5. Figure 2b shows the state after the
BDS sent seven (=#nodes - #Mirai) white-hat worms to all the nodes that do not have Mirai bots, i.e. the
6th, 7th, …, 12th nodes.
132
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
Figure 2. An application example of the all-out strategy. (a) State when the BDS detected a Mirai botnet.
(b) State after the BDS sent white-hat worms.
When the BDS adopts the All-Out strategy, the built white-hat botnet is the largest and shows the
maximum effect against the Mirai botnet. Therefore, the All-Out strategy becomes a baseline in strategy
studies. The effect of the All-Out strategy means the upper limit. The other strategies can be evaluated
depending on how close their effect is to that of the All-Out strategy.
w-Elite Strategy
The white-hat worms autonomously spread to fight against Mirai. This spread mainly depends on the
white-hat worms’ capability. If the capability is weak, the worms would die out. If the capability is too
strong, the worms would continue to stay on the system even after exterminating Mirai and waste the
resources. That is, the white-hat worms are a so-called double-edged sword.
The Few-Elite strategy Few− Elite is to launch the minimum necessary number of white-hat worms
considering the worms’ capability. It is defined as a mapping such that
Few Elite : RMirai , , #nodes , N topology , N density , , #elite if (3)
where #elite is the number of the white-hat worms to launch when the worms’ capability is sufficient, 𝛼
is a weight coefficient and 𝜃 is a threshold.
Figure 3 shows an application example of the Few-Elite strategy to the IoT system of Figure 1. Let
us assume in this example that #elite=3, 𝛼=4, and 𝜃=6. Assume that the BDS possesses two types of
white-hat worms. Type A has a capability (Lifespan ℓA=1 step, secondary infection possibility 𝜌A=75%).
Type B has a different capability (ℓB=5 steps, 𝜌B=50%). Let us first consider Type A. Its capability does
not satisfy the precondition of the strategy because ℓA+𝛼𝜌A= 1+4×0.75= 4 𝜃=6. Consequently, the
BDS will adopts not the Few-Elite strategy but the other strategies like the All-Out strategy.
133
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
Figure 3. An application example of the few-elite strategy. (a) State when the BDS detected a Mirai bot-
net. (b) State when the BDS possesses a white-hat worm with sufficient capability and sent only three.
Next, let us consider Type B. Its capability satisfies the precondition of the strategy because ℓB+𝛼𝜌B=
5+4×0.5= 7≥𝜃=6. As a result, the BDS will adopt the Few-Elite strategy and can reduce the number of
worms to launch. What is important here, the precondition of this strategy must be designed to balance
between the worms’ capability and the effect. Only then, it can become a baseline to identify whether
an available worm has sufficient capability.
Environment-Adaptive Strategy
Bots spread through the network. This spread is affected by the network’s properties such as topology
and density. Such properties may be derived from not only physical but also virtual structures like VPN.
If a network has full mesh topology, i.e. the highest network density, the bots can access all the other
nodes to infect. In contrast, if a network has tree topology, i.e. the lowest network density, the bots can
access only at most two nodes on average. Therefore, in the lower density network, the BDS should
launch more white-hat worms even if the worms’ capability is high.
The Environment-Adaptive strategy Env − Adaptive is to launch as many white-hat worms as possible
if the network density is low. It is defined as a mapping such that
Env Adaptive : RMirai , , #nodes , N topology , N density , ,
2 (4)
#nodes # Mirai if N density
#nodes
where 2𝛽/#nodes is the lowest density of connected networks composed of #nodes nodes and 𝛽 is a weight
coefficient.
134
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
Figure 4. An application example of the environment-adaptive strategy. The network has tree topology.
(a) State when the BDS detected a Mirai botnet. (b) State after the BDS sent white-hat worms. They are
sent to all the non-bot nodes because the low-density network limits spread.
Figure 2 An application example of the All-Out strategy. (a) State when the BDS detected a Mirai
botnet. (b) State after the BDS sent white-hat worms.Figure 4 shows an application example of the En-
vironment Adaptive strategy to the IoT system whose network is not full mesh but tree topology. In this
example, let us assume 𝛽=1.2. In this network, the number #AC of actual connections between nodes is
11 (=#nodes–1=12–1). We have
2 # AC 2 11
N density 0.17.
#nodes #nodes 1 12 12 1
This value satisfies the precondition of the Environment-Adaptive strategy because Ndensity=
0.17<2𝛽/#nodes= 2×1.2/12= 0.2. Therefore, the BDS will adopt this strategy and send the white-hat
worms to all the non-bot nodes.
As another example, let us apply the Environment-Adaptive strategy to the IoT system of Figure 1.
In this network, the number #AC of actual connections between nodes is 66 (=#nodes (#nodes–1)/2). We have
Ndensity= (2×66)/(12×(12–1))=1. This value does not satisfy the precondition of the Environment-Adap-
tive strategy because Ndensity=1 2𝛽/#nodes=0.2. Consequently, the BDS will adopts not the Environment-
Adaptive strategy but the other strategies like the Few-Elite strategy.
SIMULATION EVALUATION
This section describes the evaluation of the BDS and the proposed launch strategies. The evaluation has
done through simulation of the model.
135
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
Yamaguchi, S. (2020b) has proposed to regard a battle between Mirai and white-hat worms as a multi-
agent system and to express it with agent-oriented Petri net called Petri nets in a Petri net (PN2). A PN2
is intuitively a two-layers Petri net. The lower layer Petri nets (called agent nets) represent agents. The
upper layer Petri net (called environment net) represents the environment on which the agents interact
with each other. What is important here, each token of the environment net one-to-one corresponds to one
of the agent nets. Each token can move over the environment net. It can make its copy or disappear. We
assume here that the reader is familiar with Petri nets and PN2. For more information on Petri nets, refer
to (Murata, T., 1989) and (Yamaguchi, S., Bin Ahmadon, M.A., & Ge, Q.W., 2016). For more information
on PN2, refer to (Hiraishi, K., 2001) and (Yamaguchi, S., Tanaka, H., & Bin Ahmadon, M. A., 2020).
Figure 6 shows a PN2 model representing a battle between Mirai and the white-hat worms. Figure
6a-e show the agent nets while Figure 6f shows the environment net.
Figure 6a-e respectively correspond to the 1st device, the 3rd device, Mirai, the 2nd device, and the
white-hat worm. Each agent net represents the state-transition of the corresponding agent. Let us see Figure
6a. It describes the state-transition of the 1st device. Transition t1 (drawn as ÿ) represents an infection
action. Its input places p1 (drawn as 〇) and output place p2 respectively represent the state before and
after the infection. p2 has a token (drawn as •). This means that this device is in the state immediately
after the infection. Transition t2 represents a delay. Enabled transitions are highlighted in red. Place
p3 has three output transitions. The upper cycle p1t1p2t2p3t3p4t4p5t5p1 represents the behavior as a
Mirai bot, where p1 means a non-botized state while p2, p3, p4, and p5 mean botized states. Transitions
t3 and t4 represent delays. Transition t5 represents a reboot action. To sum up, the infected device has
a delay time of three steps by the reboot. The reboot clears the Mirai bot and makes the device normal
(p1). The lower cycle p1t1p2t2p3t6p6t7p7t8p1 represents the behavior of a white-hat bot. Transition t6
represents a self-destruction action for the infected white-hat worm. Therefore, P6 and p7 mean botized
states without the presence of the worm, i.e., an immune period. The middle cycle p2t2p3t9p2 represents
a behavior by secondary infection. Transition t9 represents a secondary infection action and moves from
a Mirai’s infected state (p3) to the white-hat worm infected initial state (p2). In this example, the Mirai
bot has four states p2, p3, p4, and p5. Only in p3 of them, a white-hat worm may secondarily infect it.
Therefore, the secondary infection possibility is 25%.
Next, let us see Figure 6e. It describes the state-transition of a white-hat worm. Transition t1 repre-
sents an infection action. Place p1 forms a self-loop with t1 and possesses a token. This means that the
worm can repeatedly infect a device. Transitions t2, t3, and t4 respectively represent a reboot action, a
self-destruction action, and a secondary infection action.
The environment net of Figure 6f represents the IoT system defended by the BDS. Each place (drawn
as 〇) represents a network node. This IoT system’s network has line topology and consists of three
nodes. Every place has a token (drawn as 〇) representing a device. It means that each node possesses
one device. Furthermore, place P1 has a token representing Mirai, and place P3 has a token represent-
ing a white-hat worm. This means that the device at place P1 is infected by Mirai, the device at place
P2 is normal, and the device at place P3 is infected by the white-hat worm. Each transition (drawn as ÿ)
represents an action of one agent or interaction among agents. Enabled transitions are highlighted in red.
Let us consider transition T113. It means that Mirai at P1 infects the device at P2. It is enabled because
136
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
• Agent net representing Mirai at P1 has the enabled transition t1 labeled as m_infect; and
• Agent net representing the device at P2 has the enabled transition t1 labeled as infect.
In the same way, transition T214 is enabled. It means that the white-hat worm at P3 infects the device
at P2. Let us consider the case of firing T214. This results in a new state shown in Figure 5. P2 has a new
token representing a white-hat worm. This worm is the copy of the worm at P3 and infects the device
at place P2. In this state, T113 is no longer enabled. This means that the white-hat worm protects the
device at P2 from Mirai at P1.
Figure 5. A PN2 model representing a battle between Mirai and the white-hat worms on an IoT system
defended by the BDS.
137
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
Figure 6. The state just after transition T214 fired, i.e., the white-hat worm infected the device at P2.
T113 is no longer enabled. This means that the worm protects the device at P2 from Mirai.
SIMULATION EVALUATION
We performed an experiment to evaluate the BDS and the strategies through the simulation of the PN2
model with a tool called PN2Simulator. PN2Simulator (Nakahori, K., & Yamaguchi, S., 2017) is a tool
developed in Java for editing and simulating PN2. Figure 7 shows a screenshot of the PN2Simulator.
The left-side of the screen shows an environment net, while the right-side shows agent nets. A user can
edit those nets through direct manipulation and can execute them by playing a token game interactively.
138
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
We used RMirai as the evaluation index. The value of RMirai varies with the progress of the simulation.
Therefore, it is written as a function RMirai(t) of step number t, and set t=0 when the BDS detects a Mirai
botnet and launches the white-hat worms. RMirai(0) denotes Mirai’s infection rate when the BDS detects
the Mirai botnet. RMirai(10k) denotes Mirai’s infection rate after 10000 steps. In the same way, we define
#Mirai(t), RWhite(t), and #White(t), where RWhite and #White respectively are the white-hat worm’s infection rate
and the number of the white-hat worms.
We measured RMirai(10k) by changing the following parameters.
R Mirai
, , #nodes , N topology , N density , , #elite 10 if 4 6 (5)
139
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
2 1.2
R Mirai
, , #nodes , N topology , N density , , #nodes # Mirai ifN density
#nodes
(6)
Figure 8 illustrates how to translate a network to a PN2 model. The left side of Figure 8 shows the
network of Lattice type. Let us see the 19th node. This node connects to four nodes: the 9th, 18th, 20th,
and 29th nodes. Each colored connection corresponds to the part of the PN2 with the same color. Without
a connection, the corresponding part will be omitted in the PN2 model.
Table 1. Simulation result RMirai (10k), where RMirai (0)=20% and lattice topology.
140
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
Figure 9. Influence of white-hat worm’s capability on Mira’s infection rate in the IoT system with the
lattice topology.
The simulation result is shown in Table 1 and Table 2. They respectively show the results for the Lattice
topology network and the Tree topology network. Each cell shows the mean of RMirai(10k) for 100 trials.
First, let us discuss the case of the Lattice topology. Let us see Table 1. RMirai(10k) decreased with
the increase of the white-hat worm’s lifespan ℓ or secondary infection possibility 𝜌. Figure 9 shows the
infulence of (ℓ,𝜌) on RMirai(10k). The X, Y, Z-axes respectively indicate 𝜌, ℓ, and RMirai(10k). Regardless
of #White(0), RMirai(10k) almost monotonously decreased with the increase of #White(0). This result backs
up the All-Out strategy’s validity.
141
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
When ℓ=1, the worm’s capability does not satisfy the precondition of the Few-Elite strategy (See
Equation 5) because 1 4 ¸ 6 for any 𝜌. Consequently, the BDS adopted not the Few-Elite strat-
egy but the All-Out strategy. In contrast, when ℓ=5, the worm’s capability satisfies the precondition
except for 𝜌=0%. As a result, the BDS adopted the Few-Elite strategy and can reduce #White to 10 by only
increasing RMirai(10k) at most 0.36%. From the above, we can say that the Few-Elite strategy is effective
for the Lattice topology.
Next, let us consider the case of the Tree topology. Let us see Table 2. As with the case of the Lattice
topology, RMirai(10k) decreased with the increase of ℓ, 𝜌, or #White(0). Figure 10 shows the infulence of ℓ, 𝜌,
and #White(0) on RMirai(10k). The trend is simliar but the reduction rate of RMirai(10k) siginificantly reduced
in comparison with the result of Figure 9. This is because a low-density network restrains the white-hat
142
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
worm’s spread. This network satisfies the precondition of the Environment-Adaptive strategy (See Equa-
tion 6) because Ndensity= 2.0%<2×1.2/#nodes=2.4%. As a result, the BDS adopted the Environment-Adaptive
strategy and did not reduce #White(0). If reducing #White(0) to 10, the BDS will not exterminate Mirai well.
From the above, we can say that the Enviroment-Adaptive strategy is reasonable for the Tree topology.
Figure 10. Influence of white-hat worm’s capability on Mira’s infection rate in the IoT system with the
tree topology.
Figure 11 shows the Lattice topology IoT system’s state after 10000 steps, where (RMirai, 𝛿, (#nodes,
Ntopology, Ndensity), (ℓ, 𝜌)) = (20%, 11 steps, (100, Lattice, 3.6%), (ℓ, 100%)). The horizontal axis shows
ℓ. The vertical axis shows the mean of infection rate. A stacked colored bar represents different node’s
state: red for Mirai bot, dark blue for the white-hat bot together with the worm, and light blue for only
143
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
the white-hat bot without the worm. When ℓ=1, only Mirai bots were left because the white-hat worms
were too weak to exterminate Mirai bots. When ℓ=2, the BDS successfully reduced Mirai bots, and the
white-hat worms and bots almost disappeared. This is a desirable result. When ℓ≥3, Mirai bots were
exterminated but the white-hat worms and bots remained instead. In general, it is not easy to exterminate
Mirai and further delete the white-hat worms with the worm launch strategies only. To achieve this, we
should use them with C&C strategies. Yamaguchi, S. (2021) has proposed a basic C&C strategy called
Pull-Out. As to the Pull-Out strategy, refer to (Yamaguchi, S. (2021)).
Figure 11. IoT system’s state after 10000 steps, where (RMirai, 𝛿, (#nodes, Ntopology, Ndensity), (ℓ, 𝜌)) = (20%,
11 steps, (100, Lattice, 3.6%), (ℓ, 100%)).
CONCLUSION
This chapter introduced Botnet Defense System (BDS). The BDS defends an IoT system against mali-
cious botnets. The concept of the BDS is “fight fire with fire”. To realize the concept, the BDS uses
white-hat bot technology. The BDS builds a white-hat botnet on the IoT system by itself and uses it to
exterminate the malicious botnets. The white-hat botnet autonomously spreads over the IoT system and
thus drastically increases the defense ability. On the other hand, the white-hat worm defends the IoT
system against the malicious botnet but wastes the system’s resources. Therefore, the BDS should stra-
tegically use the white-hat botnet. We presented three worm launch strategies, All-Out, Few-Elite, and
Environment-Adaptive. We regard a battle between Mirai and white-hat worms as a multi-agent system
and to express it with agent-oriented Petri net PN2. We evaluated the BDS and the proposed strategies
through simulation of the model. According to the results. the Few-Elite strategy is effective for the Lat-
tice topology, and the Environment-Adaptive strategy is reasonable for the Tree topology. This chapter
should become a trigger to interest researchers and engineers in this domain.
144
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
ACKNOWLEDGMENT
REFERENCES
Alomari, E., Manickam, S., Gupta, B. B., Anbar, M., Saad, R. M., & Alsaleem, S. (2016). A survey of
botnet-based ddos flooding attacks of application layer: Detection and mitigation approaches. In Handbook
of research on modern cryptographic solutions for computer and cyber security (pp. 52–79). IGI Global.
Arghire, I. (2017). Mirai Variant Has Bitcoin Mining Capabilities. Retrieved from https://www.secu-
rityweek.com/mirai-variant-has-bitcoin-mining-capabilities
Arzamendi, P., Bing, M., & Soluk, K. (2018). The ARC of Satori. Retrieved from https://www.netscout.
com/blog/asert/arc-satori
ASERT Team. (2018). OMG - Mirai Minions are Wicked. Retrieved from https://www.netscout.com/
blog/asert/omg-mirai-minions-are-wicked
Bailey, M., Cooke, E., Jahanian, F., Xu, Y., & Karir, M. (2009). A Survey of Botnet Technology and
Defenses. In Proc. of the 2009 Cybersecurity Applications & Technology Conference for Homeland
Security (pp.299-304). Academic Press.
Ballano, M. (2015). Is there an Internet-of-Things vigilante out there? Retrieved from https://www.
symantec.com/connect/blogs/there-internet-things-vigilante-out-there
Bezerra, V. H., da Costa, V. G. T., Barbon, J., Miani, R. S., & Zarpelão, B. B. (2019). IoTDS: A One-Class
Classification Approach to Detect Botnets in Internet of Things Devices. Sensors (Basel), 19(14), 3188.
Bonderud, D. (2016). Leaked Mirai Malware Boosts IoT Insecurity Threat Level. Retrieved from https://
securityintelligence.com/news/leaked-mirai-malware-boosts-iot-insecurity-threat-level/
Ceron, J. M., Steding-Jessen, K., Hoepers, C., Granville, L. Z., & Margi, C. B. (2019). Improving IoT
Botnet Investigation Using an Adaptive Network Layer. Sensors (Basel), 19(3), 727.
Cimpanu, C. (2018). New Mirai Variant Focuses on Turning IoT Devices into Proxy Servers. Retrieved
from https://www.bleepingcomputer.com/news/security/new-mirai-variant-focuses-on-turning-iot-
devices-into-proxy-servers/
Cisco. (2020). Cisco Annual Internet Report (2018–2023) White Paper. Retrieved from https://www.cisco.
com/c/en/us/solutions/collateral/executive-perspectives/annual-internet-report/white-paper-c11-741490.
html
Devry, J. (2016). Mirai Botnet Infects Devices in 164 Countries. Retrieved from https://www.cyberse-
curity-insiders.com/mirai-botnet-infects-devices-in-164-countries/
Edwards, S., & Profetis, I. (2016). Hajime: Analysis of a Decentralized Internet Worm for IoT Devices.
http://security.rapiditynetworks.com/publications/2016-10-16/Hajime.pdf
145
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
Frank, C., Nance, C., Jarocki, S., & Pauli, W. E. (2018). Protecting IoT from Mirai botnets; IoT device
hardening. Journal of Information Systems Applied Research, 11(2), 33–44.
Gopal, T. S., Meerolla, M., Jyostna, G., Eswari, L., Reddy, P., & Magesh, E. (2018). Mitigating Mirai
Malware Spreading in IoT Environment. In Proc. of ICACCI 2018 (pp.2226-2230). Academic Press.
Hiraishi, K. (2001). A Petri-net-based model for the mathematical analysis of multi-agent systems. IEICE
Trans. on Fundamentals, E84-A(11), 2829–2837.
Jaramillo, L. E. S. (2018). Malware Detection and Mitigation Techniques: Lessons Learned from Mirai
DDOS Attack. Journal of Information Systems Engineering & Management, 3(3), 19.
Kolias, C., Kambourakis, G., Stavrou, A., & Voas, J. (2017). DDoS in the IoT: Mirai and other botnets.
IEEE Computer, 50(7), 80–84. doi:10.1109/MC.2017.201
Manso, P., Moura, J., & Serrão, C. (2019). SDN-Based Intrusion Detection System for Early Detection
and Mitigation of DDoS Attacks. Information, 10, 106.
Meidan, Y., Bohadana, M., Mathov, Y., Mirsky, Y., Shabtai, A., Breitenbacher, D., & Elovici, Y. (2018).
N-BaIoT - Network-Based Detection of IoT Botnet Attacks Using Deep Autoencoders. IEEE Pervasive
Computing, 17(3), 12–22.
Milić, J. (2019). Mirai Botnet Continues to Plague IoT Space. Retrieved from https://blog.reversinglabs.
com/blog/mirai-botnet-continues-to-plague-iot-space
Moffitt, T. (2016). Source Code for Mirai IoT Malware Released. Retrieved from https://www.webroot.
com/blog/2016/10/10/source-code-Mirai-iot-malware-released/
Molesky, M. J., & Cameron, E. A. (2019). Internet of Things: An Analysis and Proposal of White Worm
Technology. In Proc. of IEEE ICCE 2019. Academic Press.
Murata, T. (1989). Petri nets: Properties, analysis and applications. Proceedings of the IEEE, 77(4),
541–580.
Nakahori, K., & Yamaguchi, S. (2017). A support tool to design IoT services with NuSMV. In Proc. of
IEEE ICCE 2017 (pp.84–87). IEEE.
Nakao, K. (2018). Proactive cyber security response by utilizing passive monitoring technologies. In
Proc. of IEEE ICCE 2018 (p. 1). 10.1109/ICCE.2018.8326061
netlab. (2017). Warning: Satori, a Mirai Branch Is Spreading in Worm Style on Port 37215 and 52869.
Retrieved from https://blog.netlab.360.com/warning-satori-a-new-mirai-variant-is-spreading-in-worm-
style-on-port-37215-and-52869-en/
O’Brien, S. A. (2016). Widespread cyberattack takes down sites world wide. Retrieved from https://
money.cnn.com/2016/10/21/technology/ddos-attack-popular-sites/index.html
Sinaović, H., & Mrdovic, S. (2017). Analysis of Mirai malicious software. In Proc. of SoftCOM 2017
(pp. 1-5). HR.
146
Botnet Defense System and White-Hat Worm Launch Strategy in IoT Network
US-CERT. (2016). Heightened DDoS threat posed by Mirai and other botnets. Retrieved from https://
www.us-cert.gov/ncas/alerts/TA16-288A
Yamaguchi, S. (2020a). Botnet Defense System: Concept, Design, and Basic Strategy. Information, 11, 516.
Yamaguchi, S. (2020b). White-Hat Worm to Fight Malware and Its Evaluation by Agent-Oriented Petri
Nets. Sensors (Basel), 20, 556.
Yamaguchi, S. (2021). A Basic Command and Control Strategy in Botnet Defense System. In Proc. of
IEEE ICCE 2021. Academic Press.
Yamaguchi, S., Bin Ahmadon, M. A., & Ge, Q. W. (2016). Introduction of Petri Nets: Its Applications
and Security Challenges. In B. Gupta, D. P. Agrawal, & S. Yamaguchi (Eds.), Handbook of Research
on Modern Cryptographic Solutions for Computer and Cyber Security (pp. 145–179). IGI Publishing.
Yamaguchi, S., & Gupta, B. (2019). Malware Threat in Internet of Things and Its Mitigation Analysis.
In R. C. Joshi, B. Gupta, D. P. Agrawal, & S. Yamaguchi (Eds.), Security, Privacy, and Forensics Issues
in Big Data (pp. 363–379). IGI Publishing.
Yamaguchi, S., Tanaka, H., & Bin Ahmadon, M. A. (2020). Modeling and Evaluation of Mitigation
Methods against IoT Malware Mirai with Agent-Oriented Petri Net PN2. International Journal of Internet
of Things and Cyber-Assurance, 1(3/4), 195–213.
147
148
Chapter 9
A Survey on Emerging
Security Issues, Challenges,
and Solutions for Internet
of Things (IoTs)
Anish Khan
UIET, Kurukshetra University, Kurukshetra, India
Dragan Peraković
https://orcid.org/0000-0002-0476-9373
University of Zagreb, Croatia
ABSTRACT
The internet of things is a cutting-edge technology that is vulnerable to all sorts of fictitious solutions.
As a new phase of computing emerges in the digital world, it intends to produce a huge number of smart
gadgets that can host a wide range of applications and operations. IoT gadgets are a perfect target for
cyber assaults because of their wide dispersion, availability/accessibility, and top-notch computing
power. Furthermore, as numerous IoT devices gather and investigate private data, they become a gold
mine for hostile actors. Hence, the matter of fact is that security, particularly the potential to diagnose
compromised nodes, as well as the collection and preservation of testimony of an attack or illegal activ-
ity, have become top priorities. This chapter delves into the timeline and the most challenging security
and privacy issues that exist in the present scenario. In addition to this, some open issues and future
research directions are also discussed.
INTRODUCTION
Since from the beginning till the present day there are numerous definitions coined by different authors
and researchers. The term “Internet of Things” refers to a network of physical devices or embedded
devices which are capable of sending and receiving information or data via internet know as “Internet
of Things (IoT)” (Oracevic et al., 2017, Adat et al., 2018, Lv, Z., 2020, Liu et al., 2017), Khattak et al.,
DOI: 10.4018/978-1-7998-7789-9.ch009
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
2019). A technology that combines real-world with virtual world offers a rich ground for imagination
and innovative ideas. The IoT is next phase in evolution of communication technologies. We caught a
glimpse in the manner of utilization of technologies is an eye catching transformation that moulds the
world from off the beaten track systems to ubiquitous Internet-enabled “things” (Sain et al., 2017).
For trouble free identification and communication purposes, digital identities are issued to every
smart object that will help in sharing data and opportunity to use various services. A unique detection-
system-radio frequency identification (RFID) has been developed using the concept of digitally identi-
fying many gadgets (Aldowah et al., 2018, Aly et al., 2019, Sahmim & Gharsellaoui, 2017, Arıs et al.,
2018). Wireless sensor networks are crucial in the execution of cutting-edge technologies because they
are resource constrained. The use of cloud computing results in the creation of a virtual platform for
integrating storage devices, development and research tools, and so on (Jose, D. V., & Vijyalakshmi, A.,
2018). Users can access applications on demand without considering their physical location. As IoT is
a vast area and offers number of applications in every possible way in routine life. For example, health
care monitoring systems, smart transportation systems, retail, surveillance, wearable gadgets and many
more (Jose, D. V., & Vijyalakshmi, A., 2018, Alaba et al., 2017, Bhattarai, S., & Wang, Y., 2018). Figure
1: Illustrate some of the common applications of Internet of things.
As far as IoT becomes burning matters nowadays, in contrast with this many security and privacy
challenges have pop-up and become the bottle neck issues for IoT as the number of edge devices has
increased dramatically (Yousefi, A., & Jameii, S. M., 2017). Various real-time attacks, such as zero-day
attacks, ransomware, phishing attacks, and DDoS attacks (Sadeeq et al., 2018, Chahid et al., 2017, Has-
san, W. H., 2019), have been introduced in recent years. The motive of this paper is to highlight various
attacks that make edge devices vulnerable.
The road map of this paper contains eight sections. Section I gives a short and crisp introduction to
IoT. Section II tells us about history and growth of IoT. Section III highlights the main elements of IoT.
Section IV illustrates the three layer architecture of Internet of Things. Section V gives brief introduction
of taxonomy of attacks. Section VI gives bird eye view of security challenges to each layer of IoT and
Section VII illustrates some counter measures of various threats. Section VIII tells some future research
directions and some open challenges to IoT. Finally, Section VIII concludes this paper.
TIMELINE OF IoT
The “Internet” is a worldwide group of combined servers, PCs, tablets and mobiles that are administered
by standard protocols for combined frameworks. This enables users to send, receive and communicate
the information (Sfar et al, 2018, Ande et al., 2020, Conti et al., 2018, Maple et al., 2017). The word
“Things” has many meanings in English dictionary. The word thing refers to an object, action and situ-
ation. For example, a mobile is referred to an object, ‘those kinds of things are expected from her’- here
things are referred as action (Li, S., & Da Xu, L., 2017).
With the combination of above mentioned terms, a new term originates called “Internet of Things”
that means a network of physical devices or embedded devices which are capable of sending and receiv-
ing information or data with the help of internet know as “Internet of Things” (Madakam et al., 2015).
Vision of Internet of Things is to make things (tube-light, AC, fan, door bell, table etc) smart and to act
like living entities by using internet. Internet becomes ubiquitous and spread almost in every part of the
world and human life is directly influenced by internet.
149
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Moreover Internet of Things has no fixed definition; many definitions are available in the literature.
Let’s have a brief view of some definitions that are available in the literature, Mark-Weiser gave a central
statement in seminal paper “The most profound technologies are those that disappear. They weave them-
selves into the fabric of everyday life until they are indistinguishable from it” at Scientific American in
1991 (Mattern et al., 2010). An expert on digital innovation named Kevin Ashton stated “An open and
comprehensive network of intelligent objects that have the capacity to auto-organize, share information,
data and resources, reacting and acting in face of situations and changes in the environment” in 1998
(Mattern et al., 2015).
In layman’s terms, the IoT is a framework of interconnected gadgets. It leverages the Internet to
transfer information and practically any gadget we may think about can be attached to it. In reality, the
factors that fueled the rise of smart consumer electronics, such as ubiquitous wireless communication,
cloud computing, low-cost sensors, and improved AI, are now being utilized in tandem with big data to
strength the next age of industry (Sun et al,2018, Hou et al., 2019, Lin et al., 2017). With the passage of
time much advancement are being done in the field of embedded systems and nano-electronics devices
that makes possible in the reduction of size and dramatic improvements in computational process. These
nano-electro devices are now available with global positioning systems (GPS), attached with the things
to make them visible and provide a real time monitoring system (Grammatikis et al., 2019, Al-Garadi
et al., 2020). Figure 2 give us a crisp overview of timeline of IoT.
In the month of March 2020, Cisco Annual Internet Report was published that analysis various segments
like, small scale business, private sector, enterprises etc. The report envelopes quantitative predictions
and calculations in respect to expansion of internet users, devices and network execution and capabilities.
150
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
151
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Figure 3 is illustrating growth in the number of users from the past year 2018 to the future 2023. It
can be derived from bar graph that number of users will increases up to 2/3 of total population and will
reach to the count of 5.3 billion till 2023. In addition to this, IoT applications will take over the market,
the estimated market share will increases up to 1567 billion by 2025 onwards. Figure 4. tells about the
dramatic expansion in the market size.
152
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
However, with the rapid growth in IoT technology approximately 66% of total population will have
access to the internet. There is an unbelievable growth in market shares. The broadband speed will boost
up to 110.4 Mbps, WiFi will gain momentum and reaches at a peak of 628 million hotspots. But the fact
to be considered is that the risk of security will remain increased with the improvement in the technol-
ogy. The number of attacks on the machine will almost double as compared to the present scenario. In
the year 2018, the number of DDoS attacks was 7.9 million but it can be seen clearly that in the coming
years (2023) it will reach at 15.4 millions. Figure 5: gives a bird eye view of expansion in number of
DDoS attacks.
IoT ELEMENTS
Users can take advantage of a variety of advantages and features provided by the Internet of Things.
As a result, specific elements are required in order to use them properly. The components required to
provide IoT functionalities (Mendez Mena et al., 2018, Samaila et al., 2017, Sha et al., 2018, Misra et
al., 2017, Ge et al., 2017, Razzaq et al., 2017, Bertino & Islam, 2017, Dabbagh, & Rayes, 2019, Butun
et al. (2019), HaddadPajouh et al., 2019, Gulzar & Abbas, 2019, Yousuf & Mir, 2019). These elements’
names and descriptions are as follows:
153
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
for naming network devices, including electron product codes (EPC) and ubiquitous codes. IPv6
is used to provide each object a unique address. Initially, IPv4 was utilized to allocate addresses,
but owing to the rising number of IoT devices, it was unable to meet the demand for addressing
(Samaila et al., 2017, Misra et al., 2017).
2. Sensing: The sensing is the process of gathering information from objects. The information gath-
ered is transmitted to the depository. Actuators, RFID tags, sensor technologies, wearable sensing
devices, and other sensing devices are used to gather data from surroundings and devices (Sha et
al., 2018, Ge et al., 2017).
3. Communication: The principal objective of the Internet is to connect and impart amongst several
gadgets. Devices can transmit and accept delivery of information, files, and other data during com-
munication. Many technologies, including as Near Field Communication (NFC), Bluetooth, Wi-Fi,
Long Term Evolution (LTE), and Radio Frequency Identification (RFID), enable communication
(Razzaq et al., 2017, Bertino, & Islam, 2017).
4. Computation: Computing performs on the information collected from devices by means of sen-
sors. Numerous hardware and software frameworks have been created to handle computation in
IoT applications. Audrino, Raspberry Pi, are utilised for hardware platforms, while the OS is used
154
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
for software platforms to execute the processing. Tiny OS, Lite OS, Android, and other operating
systems are among the options (Dabbagh, & Rayes, 2019, HaddadPajouh et al., 2019, Gulzar &
Abbas, 2019).
5. Service: There are four sort of assistance provided by Internet of Things applications. The first is an
identity management service. It is in the process of examined which devices made a call. Another
service that collects whole information from devices is data gathering. The second last service
makes judgments based on the data obtained and responds appropriately to the gadgets. Fourth
and final service is universal service, which is utilised to reply to devices instantly, regardless of
time or location (Butun et al., 2019, HaddadPajouh et al., 2019).
6. Semantic: It is the primary duty of the Internet of Things (IoT) to assist people by completing
their duties. To fulfil its obligations, it is the most crucial aspect of IoT. It functions as the IoT’s
brain. It gathers all relevant data and put together a favourable judgment about how to respond to
the devices (Gulzar et al., 2019, Yousuf, & Mir, 2019).
IoT ARCHITECTURE
IOT is not an individual technology; it is a cluster of different technologies. Hence, to implement this
technology, various IOT architectures are available in literature proposed by different researchers. Here
we only discuss the three layered architecture for IoT applications. The layers are “Perception layer”,
“Network Layer” and “Application layer” (Das et al., 2018, Virat et al., 2018, Yang et al., 2017, Salam,
2020, Husamuddin & Qayyum, 2017, Vashi et al., 2017, Chen et al., 2018, Ammar et al., 2018, Liu et
al., 2017, Yaqoob et al., 2017, Mendez et al., 2017, Hameed et al., 2019, Rizvi et al., 2018, Gaurav et
al., 2021) as illustrated in Table 1.
1. Perception Layer: This layer is often called “Device Layer”. The IoT architecture’s lowest layer
is referred to as “Sensory Layer” or “Recognition Layer”, which comprise of sensors that are re-
sponsible for the sensing and collecting data from sensors (Das et al., 2018, Virat et al., 2018, Yang
et al., 2017). It involves sensing (gathering information from the environment and delivering it to
database servers, or the cloud), and recognition methods (recognize devices on the basis of unique
identity assigned to them), with minimal human engagement, actuation (taking a mechanical ac-
tion based on sensed data) and communication (creating connectivity among heterogeneous smart
devices) are possible (Singh & Vardhan, 2019). It is distinguished by the ability to capture data from
the actual environment and display it in a digital representation. This perception layer can be sub
categorised into two categories i.e. Perception Nodes and Perception Network. Perception Nodes
contains the Sensors which are utilized to perform an action of information gathering from their
respective surroundings. Sensors can be Radio Frequency Identification tag (RFID tags), Barcodes,
Temperature sensors, humidity sensors, Near Field Communication sensors etc (Salam, 2020).
After gathering information from sensors, it is forwarded to the next layer i.e. Network Layer. The
modem through which gathered information is transferred to the next layer is Perception Network
(Das et al., 2018, Husamuddin & Qayyum, 2017), which is responsible for the transferring data
with wired or wireless communication modems.
155
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Access Layer
Information,
Network Layer Core Network Connectivity and WiFi, Internet, GPRS
Transmission
LAN & WAN
Perception Network
Elements Technologies
Identification IPv4, IPv6, Product Code
Sensing Sensors, Wearable Devices, Actuators
Communication RFID, WSN, NFC, WiFi
Computation Arduino, Raspberry Pi, Windows OS
Service Identity Based, Data Aggregation, Shared Service and Universal Services
Semantic RDF, OWL, EXI
2. Network Layer: The “Network Layer” is often called as “Transportation Layer” and “Transmission
Layer”. It is found in between perception layer and application layer. This layer is the integration of
the various heterogeneous networks and protocols that are responsible for the transmission of the
gathered data from the perception layer. Its aim is to convey data collected via wired or wireless
communication medium for the processing, data extraction, data assembling and encryption of
perception nodes to the information processing unit (Yang et al., 2017, Husamuddin & Qayyum,
2017, Vashi et al., 2017, Hossain et al., 2019). This layer is further divided into three categories
i.e. Access, Core and Local/Wide Area Network.
a. Access Layer: It is a sort of telecommunications network that acts like a link between subscrib-
ers and service providers. It develops communication and infrastructure abilities for target
users, such as mobile communication, satellite communication, and wireless communication.
156
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
ZigBee, 2G and 3G, Adhoc, GPRS, Wi-Fi, Low Energy Bluetooth networks are examples of
access networks that IoT can use. 4G-LTE/5G are a sophisticated telecom technology that
allows mobile devices to connect to the Internet at high speeds (Das et al., 2018, Virat et al.,
2018, Husamuddin & Qayyum, 2017).
b. Core Network: The Internet serves as the IoT’s Core Network, providing the necessary in-
frastructure. It is responsible for pass on information among connected clients via network
access. Moreover it serves as backbone of every communications system and functioning as
a gateway for data exchange. It facilitates resource pooling communication among restricted
gadgets. It allows us to monitor and control physical items from a distance (Vashi et al., 2017,
Chen et al., 2018, Ammar et al., 2018).
c. LAN/WAN Networks: A local area network associate gadgets in a very restricted area. LAN
devices may interact directly with one other and can use gateways to communicate with
distant devices. Similarly, Wide Area Networks (WANs) are considered to distribute devices
over wider regions. Low-power WAN’s is capturing popularity for the reason that they allow
low-power gadgets to communicate with each other (Chen et al., 2018, Liu et al, 2017).
3. Application Layer: Application Layer is uppermost layer in three layered architecture that is per-
ceivable to users. The application layer’s objective is to supervise and attire applications worldwide
using data gathered from the perception layer, which is analyzed by the IT unit. It offers access by
using various portable and terminal devices to the end users across the network according to their
needs. Application Support Layer and IoT Applications are the two sub layers that make up this
layer (Yaqoob et al. 2017, Mendez et al. 2017).
a. Application Support Layer: It supports a wide range of corporate services and is in charge
of conducting intelligent calculations and data processing. It is in charge of executing smart
calculations and data processing and supports a diverse variety of company services. It utilizes
middleware, which comprises of workstations for software installment on many devices and
applications and is accountable for sophisticated operations. Machine-to-machine applications
allow systems to communicate directly with one another via wired or wireless connections
(Hameed et al. 2019, Rizvi et al. 2018, Sial et al. 2018).
b. IoT Applications: Internet of things is now an important part of our life. Every day we are fa-
miliar with many smart device moreover we are using many smart devices in our daily routine.
Internet has made everything easy and information is available on our figure tips (Makhdoom
et al. 2018, Nurse et al. 2017, Miloslavskaya & Tolstoy 2019). As assumed in late 1990’s that
till 2020, billions of users will be connected to the internet and this assumptions came true.
Internet of things offers number of applications to make life much easier and hassle free like
smart home, smart city, smart transport system and many more as illustrated in figure 1.
With IoT’s economic, social and technological importance increasing every day, IoT not only attracts
investment, user and scientist but also malicious users who work hard and find a way to make linked
devices cyber attack weapons or to steal data. Although it is apparent that linking a huge number of
non-conventional equipment via the Internet, including water supply pumps and intelligent appliances,
is a scary idea behind the establishment of IoT, and will bring about drastic changes to security and
157
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
privacy for people and businesses. For example, Proof point security investigators have uncovered cyber
assaults that could be the first IoT Botnet attack to exploit around one lakh intelligent domestic appli-
ances like Smart Televisions, smart lighting, Smart Drainage systems, and other appliances. Between
the time periods 23rd Dec 2013 till 6th Jan 2014, Around 750,000 malicious spam emails were sent to a
range of enterprises and people around the world by attackers (Tzafestas, 2018, Jindal et al. 2018, Sial
et al. 2019, Van Oorschot & Smith, 2019).
Although the issues related to security and privacy are not new. Because of the increased IoT popular-
ity more and more devices are deployed or integrated into different systems, organizations, houses, etc.
that collect all types of information. The majority of these devices include a variety of security flaws that
might be used by hostile actors to damage authorized customers. Various consumers may be unaware
that they are being filmed or monitored due to the abilities of several smart gadgets.
The way that information is quickly accessible by means of the Internet and may conceivably be gotten
to through ineffectively ensured keen gadgets with practically no safety efforts is one empowering angle
that will make IoT information a really engaging objective for criminal clients. As a result, it’s critical
to make sure that each device is adequately protected. By addressing these issues and guaranteeing IoT
security, attackers will be prevented from exploiting vulnerabilities that might lead to the compromis-
ing of confidential information and sensitive data. The concept of “one-size-fits-all” criteria cannot be
efficiently employed with every specific application inside the IoT environment. The following core
security properties can be used to classify security threats: “confidentiality”, “integrity”, “availability”
(CIA), “authentication”, “access control”, “non-repudiation”, “secure booting”, and “device tampering”
detection (Salman & Jain, 2019, Alrawais et al. 2017, Saha et al. 2017, Mastorakis et al. 2017, Mavro-
poulos et al. 2019). Table 3 shows various kinds of attacks present in the current scenario.
Denial-of-Service (Dos),
These are typically used to take out hostile activities in opposition to
Distributed Denial of Service (DDOS),
Active Attacks the system, damaging or interrupting legitimate users’ services. They
Man-in-the-Middle (MitM), Interruption,
threaten the system’s secrecy and integrity.
Alteration
These are mostly used to collect relevant information without being Monitoring, Traffic Analysis,
Passive Attacks
detected, i.e., they do not interfere with conversation. Eavesdropping, Node destruction/malfunction
The features of the Internet of Things allow it to conduct targeted Identification, profiling, tracking, linkage,
Privacy Threats
attacks on users’ privacy. inventory
Protocol-Based Attacks The attacks target Internet of Things (IoT) connection standards. RFID-based, Bluetooth based, Zigbee Based
158
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
As illustrated in Table 4, different security concerns that compromise anonymity on each layer that have
recently been presented are briefly reviewed in this section.
1. Perception Layer: The physical layer is made up of numerous enabling sensor technologies includ-
ing Bluetooth, GPS, and Zigbee, all of which are vulnerable to various sorts of intrusions. These
sensors may capture data regarding location, air quality, surroundings, activity, and vibration etc.
The main aim of the attacker is to exploit those sensors and to replace those with its own malicious
sensors. The attacks listed below are carried out on the hardware components of the IoT network,
and the attacker must be proximate to the IoT systems (Bhatt & Ragiri, 2021, Lee & Kim, 2017,
Yu et al. 2018, Tewari & Gupta, 2020, Patwary et al. 2020). Conventional security dangers and
issues to perception layers are:
a. Eavesdropping: Eavesdropping is an illegal real-time attack in which an attacker intercepts
confidential conversations such as telephone conversations, text messages, fax transfers, or
video chat. It aims to retrieve information that is being sent via a network (Lee & Kim, 2017).
It utilizes insecure communication to gain access to the data being delivered and processed.
b. Node Capture: It is among the most dangerous assaults in the IoT’s perception layer. A critical
node, such as a gateway node, is taken over by an attacker. It has the potential to leak any data,
including conversation between sender and recipient, a key utilized for secure connection,
and data kept in storage (Yu et al. 2018).
c. Malicious Node: In this kind of attack, an assailant installs a malicious node and misrepresents
the information. Its primary aim is to halt the channeling of accurate data (Tewari & Gupta,
2020). A hacker inserts a node to devour precious energy from genuine nodes and successfully
govern to damage the network.
d. Replay Attack: A playback attacks is another name for it. It is an assault by an intruder that
transmits and receives legitimate information on the conservation of the sender (Ali et al.
159
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
2019). By demonstrating evidence of his identity and authenticity, an intruder provides the
identical verified data to the target that was already received in his conversation. Because the
communication is encrypted, the recipient might consider it as a legitimate request and take
the attacker’s specific behavior (Gupta & Quamara, 2020).
e. Timing Attack: This attack often seen in gadgets with restricted computational capability.
It is nothing but an aggressor to discover defects in a framework’s uprightness and recover
mysteries by observing what amount of time it will require for the framework to answer to
different solicitations, input, or cryptographic techniques (Gubbi et al. 2013).
2. Network Layer: It is also known as “Transportation Layer”. It is found in between perception layer
and application layer. It also sets out the framework for interconnecting smart gadgets with network
infrastructure. It poses serious security concerns with the integrity and authenticity of data being
transmitted across the network Conventional security dangers and issues to network layer are (Ren
et al. 2017, Chegini & Mahanti, 2019):
a. Denial of Services (DoS) Attack: A denial-of-service attack (DoS) is a type of cyber-attack
that prevents legitimate users from retrieving gadgets or other network resources (Al-Garadi
et al. 2020). It’s usually done by sending repeated requests to the targeted devices or network
resources, making it extremely difficult or inconvenient for some or all legitimate users to
utilize them.
b. Man-in-the-Middle Attack (MitM): A MiTM attack occurs when an attacker privately en-
crypts and changes information between a transmitter and a recipient who believe they are
speaking directly with each other (Kouicem et al. 2018, Ogonji et al. 2020, Madakam et al.
2015, Mattern, F., & Floerkemeier, C. 2010, u Farooq et al. 2015). Because the intruders have
authority over the communication, he or she may modify messages to suit their purposes. It
poses a significant danger to internet security since it allows an attacker to capture and alter
data in real time (Roman et al. 2011).
c. Storage Attack: Users private data is saved on storage devices or on the cloud. The attacker
can attack both storage devices and the cloud and user details may be altered (Ray, 2018).
The duplication by various sorts of individuals of information connected with access gives
more opportunities for exploitation (Stankovic, 2014).
d. Exploit Attack: Any unethical or unlawful assault as programming, bits of information, or a
string of instructions is referred to as exploitation. In an application framework or equipment,
it uses the security vulnerabilities (McEwen & Cassimally, 2013, Madakam et al. 2015, Chen,
2012, Al-Qerem et al. 2020). This is generally used to manipulate the system and capture data
stored on a network.
3. Application Layer: All applications that employ IoT technology or IoT have been implemented
are defined by the application layer. “Smart homes, smart cities, smart health, animal tracking”,
and other IoT applications becoming feasible (Stergiou et al. 2020, Tewari & Gupta 2017, Cvitić et
al. 2021, Masud et al. 2020, Salhi et al. 2021). It is responsible for providing services to the apps.
a. Phishing attack: It’s a delicate social engineering tactic that involves impersonating a trusted
entity in order to get access to user credentials such as passwords and credit card numbers
(Masud et al. 2020, Salhi et al. 2021, Hallappanavar, V. L., & Birje, M. N., 2021, Mishra et
al.,2021, Dahiya & Gupta, 2021, Bhushan & Gupta, 2019).
160
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
b. Session hijacking attack: The hacker takes advantage of the user’s online session to obtain ac-
cess to the user’s sensitive data (Chhabra et al. 2013, Mirsadeghi et al. 2020, Gou, Yamaguchi
& Gupta, 2017, Tewari & Gupta, 2020, Gupta & Quamara, 2020).
c. Cross Site Scripting: It’s a kind of code infusion assault (Stergiou et al. 2020, Wei et al. 2016,
Arias et al.,2015), Harbi et al. 2019). It’s nothing but an assailant to infuse a customer side
content, for example, java script, onto a confided in site that different clients are visiting. An
assailant can altogether change the design and usefulness of the program to suit his requests
and use unique information in an unlawful way along these lines (Khattak et al. 2019, Nandy
et al. 2019).
d. Malicious Code Attack: It’s a piece of programming code that is intended to have potentially
negative results and obliterate the framework (Tahsien et al. 2020, Elhoseny et al. 2020). It is
a sort of a kind of malware that anti-virus will not have the option to stop or oversee. It can
either fire up all alone or work as a program that involves the user’s observations to complete
a task (Aldowah et al. 2018, Aly et al. 2019, Arıs et al. 2018).
Although with the rapid increase in cyber attacks many mechanisms are designed and proposed to tackle
with disastrous attacks. Taxonomy of some of the counter measures are illustrated below in figure 7:
161
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
In Perception layer below mentioned techniques are implemented to tackle attacks that occur in
particular layer:
1. Hash Based Encryption: To assure users information to be secure, hash based encryption technique
is came into existence, it converts the information into unknown form called “Cipher Text”. When
information is shared among authentic user sender encrypt the data with a key and only authorized
receiver can access the information with the decryption key (Li & Xiong, 2013, Sundaram et al.
2015).
2. Public Key Infrastructure: The concept of public key and private key are implemented here. The
public key is stored at the base station and the private is distributed among all the nodes that are
present in the IoT network. Shared message is encrypted and decrypted with the help of private
key at receivers end (Weber, 2010, Li et al. 2013).
3. Device Authentication: Without any authentication the device cannot enters or connect with other
nodes in the IoT network (Airehrour et al. 2017).
4. Physically Secure Design: Designing of physically secure devices should be of high quality (Suto,
2010).
In Network Layer below mentioned techniques are implemented to tackle attacks that occur in par-
ticular layer:
1. SDN: It stands for Software Defined Network; it provides better results and performance at low
cost and uses less network resources. The integration of SDN architecture and IoT architecture are
utilized for better security purposes. Combined architecture contains only three devices that are:
IoT Agent, IoT Controller and SDN Controller (Robertazzi, 2017, Al Shuhaimi et al. 2016).
2. Ad-hoc Routing: It stops inside attacks from the network (Brakerski & Vaikuntanathan, 2014).
3. Hello Flood Detection: In this scenario, a node sends hello message to check the strength of the
signal. If the strength of the signal is similar to the radio range then receiver accepts the message
(Ongtang et al. 2012).
In Application Layer below mentioned techniques are implemented to tackle attacks that occur in
particular layer:
1. Virtual Identity Framework: In the IoT scenario, user’s information is being collected like health,
home address, mobile number etc, for processing and transmission. It is the top priority to secure
users identity and personal information from unauthorized and attackers. The users just request the
service provider for their VID. The service provider asks some of the queries like age, name, sex
and D.O.B and provide them a VID. The information of the users is now accessed with the help
of virtual ID (Hu et al. 2011).
2. Access Control Mechanism: While establishing the network for IoT, there is a huge number of
devices that are connected to each other. For their seamless connection every device requires a
unique Id for the communication. With the increase in number of devices, the risk of attack is
also increased. Attacks like, Denial of Service (DoS) attack, Man-in-the-middle attack came into
existence, to overcome these issues the role of authorization and access control techniques are
implemented (Bormann et al. 201, Gupta, K & Shukla, 2016).
162
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
3. Identity Based Security Systems: These kinds of systems are very crucial for the detection of user’s
location in emergencies. Only the authorized person can have access to the live location. The system
contains 4 subsystems named as, Client subsystem, Registration subsystem, Users Policy subsystem
and User Authentication subsystem (Hu et al. 2011, Sarma et al. 2008).
1. Standardization: Considering most of the studies conducted so far, there’s been no universal stan-
dardization for the IoT anatomy (Jose & Vijyalakshmi, 2018, Alaba et al. 2017, Florea et al. 2017,
Obaidat et al. 2019, Li et al. 2019, Yousefi & Jameii, 2017) [110-115]. The evolution of complex
ideas has been sparked by various security challenges and application requirements (Sklavos &
Zaharakis, 2016, Gelenbe et al. 2018, Kalyani & Chaudhari 2020, Narendrakumar et al. 2018).
However, it has become an epic match to integrate the objectives and features with one full-fledged
model for multiple applications.
2. Security and Privacy: An enormous security and privacy threats will appear when IoT networks
are deployed on a massive scale. Smart gadgets are inadequate to prop up with extremely secure
protocols and algorithms due to resource limitations (Li et al. 2016, Wang et al. 2017, Hassan,
2019, Hossain et al. 2015, Andrea et al. 2015). Clients must always be informed of who will be
acquiring their private data, how it is being gathered, and how it will be implemented in order to
maintain their virtual privacy (Krichen et al. 2017, Sfar et al. 2018, Kolias et al. 201).
3. Power and Energy Efficient: Using various heterogeneous gadgets for sensing information from
real world, bandwidth congestion across communication channels, data computation, storage ex-
haustion and energy depletion of devices are directly affected. To overcome these issues, energy
harvesting systems (Mahmoud et al. 2015, Kouicem et al. 2018, Li et al. 2015, Maple, C. 2017, Li
et al. 2017, Granjal et al. 2015, Sun et al. (2018)[128-134].
4. Data Mining/ Big Data: Building connectivity between millions of devices via the Internet will
produce enormous amounts of data, so termed as big data. There is inadequacy in dealing with
operational network architectures and software applications (Hou et al, 2019, Lin et al. 2017,
Vasilomanolakis et al. 2015). For the storage, fetching and refining of the bulk data, appropriate
techniques are necessary like, data mining techniques, supervised, unsupervised and deep learn-
ing techniques (Chasaki & Mansour, 2015, Grammatikis et al. 2019, Wortmann & Flüchter, 2015,
Sicari et al. 2015).
5. Cloud Computing: Merging cloud computing with Internet of Things will accelerate the growth
of intelligent apps that can handle thousands of users and a huge number of devices (Al-Garadi
et al. 2020, Samaila et al. 2017). Applications must run in both wired and wireless environment.
Moreover, real time apps should be worked in parallel so that if any miss happening occurs there
will be no compromise with the processing (Ge et al. 2017).
6. Denial of Services (DoS) attacks: Smart devices are resource constrained that only focus on execu-
tion of the functioning rather than building a secure environment. This is the main cause behind
smart gadgets becoming vulnerable and acting like gold mines for attackers (Madakam et al. 2015,
Razzaq et al. 2017, Sadeghi et al. 2015, Tankard, C. 2015). Various newly introduced variants of
163
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
DoS attacks have come into existence that will come into consideration when the whole network
shuts down. Some well-known DoS attacks in the scenario include Mirai, Hajime, and BrickerBot
[150].
CONCLUSION
The Internet of Things is an emerging paradigm which helps us to communicate with non living objects
or things with the integration of internet. By considering the capabilities of IoT and smart gadgets that
boosts the growth in this field. Millions and billions of devices are connected with each other for the
improvement of our lifestyle. In addition to this, many malicious activities are going on that violate our
security and privacy. Here in this paper, we briefly go through trending technology and then security
challenges related to each layer of the framework. Never the less, taxonomy of various counters measures
and some security issues are discussed for the future perspective.
REFERENCES
Adat, V., & Gupta, B. B. (2018). Security in Internet of Things: Issues, challenges, taxonomy, and ar-
chitecture. Telecommunication Systems, 67(3), 423–441. doi:10.100711235-017-0345-9
Airehrour, D., Gutierrez, J., & Ray, S. K. (2017). A trust-aware RPL routing protocol to detect blackhole
and selective forwarding attacks. Journal of Telecommunications and the Digital Economy, 5(1), 50–69.
Al-Garadi, M. A., Mohamed, A., Al-Ali, A. K., Du, X., Ali, I., & Guizani, M. (2020). A survey of ma-
chine and deep learning methods for internet of things (IoT) security. IEEE Communications Surveys
and Tutorials, 22(3), 1646–1685.
Al-Garadi, M. A., Mohamed, A., Al-Ali, A. K., Du, X., Ali, I., & Guizani, M. (2020). A survey of ma-
chine and deep learning methods for internet of things (IoT) security. IEEE Communications Surveys
and Tutorials, 22(3), 1646–1685.
Al-Garadi, M. A., Mohamed, A., Al-Ali, A. K., Du, X., Ali, I., & Guizani, M. (2020). A survey of ma-
chine and deep learning methods for internet of things (IoT) security. IEEE Communications Surveys
and Tutorials, 22(3), 1646–1685.
Al-Qerem, A., Alauthman, M., Almomani, A., & Gupta, B. B. (2020). IoT transaction processing through
cooperative concurrency control on fog–cloud computing environment. Soft Computing, 24(8), 5695–5711.
Al Shuhaimi, F., Jose, M., & Singh, A. V. (2016, September). Software defined network as solution
to overcome security challenges in IoT. In 2016 5th International Conference on Reliability, Infocom
Technologies and Optimization (Trends and Future Directions)(ICRITO) (pp. 491-496). IEEE.
Alaba, F. A., Othman, M., Hashem, I. A. T., & Alotaibi, F. (2017). Internet of Things security: A survey.
Journal of Network and Computer Applications, 88, 10–28. doi:10.1016/j.jnca.2017.04.002
164
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Alaba, F. A., Othman, M., Hashem, I. A. T., & Alotaibi, F. (2017). Internet of Things security: A survey.
Journal of Network and Computer Applications, 88, 10–28.
Aldowah, H., Rehman, S. U., & Umar, I. (2018, June). Security in internet of things: issues, challenges
and solutions. In International Conference of Reliable Information and Communication Technology
(pp. 396-405). Springer.
Aldowah, H., Rehman, S. U., & Umar, I. (2018, June). Security in internet of things: issues, challenges
and solutions. In International Conference of Reliable Information and Communication Technology
(pp. 396-405). Springer.
Ali, I., Sabir, S., & Ullah, Z. (2019). Internet of things security, device authentication and access control:
a review. arXiv preprint arXiv:1901.07309.
Alrawais, A., Alhothaily, A., Hu, C., & Cheng, X. (2017). Fog computing for the internet of things:
Security and privacy issues. IEEE Internet Computing, 21(2), 34–42.
Aly, M., Khomh, F., Haoues, M., Quintero, A., & Yacout, S. (2019). Enforcing security in Internet
of Things frameworks: A systematic literature review. Internet of Things, 6, 100050. doi:10.1016/j.
iot.2019.100050
Aly, M., Khomh, F., Haoues, M., Quintero, A., & Yacout, S. (2019). Enforcing security in Internet of
Things frameworks: A systematic literature review. Internet of Things, 6, 100050.
Ammar, M., Russello, G., & Crispo, B. (2018). Internet of Things: A survey on the security of IoT
frameworks. Journal of Information Security and Applications, 38, 8–27.
Ande, R., Adebisi, B., Hammoudeh, M., & Saleem, J. (2020). Internet of Things: Evolution and tech-
nologies from a security perspective. Sustainable Cities and Society, 54, 101728.
Andrea, I., Chrysostomou, C., & Hadjichristofi, G. (2015, July). Internet of Things: Security vulnerabilities
and challenges. In 2015 IEEE symposium on computers and communication (ISCC) (pp. 180-187). IEEE.
Arias, O., Wurm, J., Hoang, K., & Jin, Y. (2015). Privacy and security in internet of things and wearable
devices. IEEE Transactions on Multi-Scale Computing Systems, 1(2), 99–109.
Arıs, A., Oktug, S. F., & Voigt, T. (2018). Security of internet of things for a reliable internet of services.
Academic Press.
Arıs, A., Oktug, S. F., & Voigt, T. (2018). Security of internet of things for a reliable internet of services.
Academic Press.
Bertino, E., & Islam, N. (2017). Botnets and internet of things security. Computer, 50(2), 76–79.
Bhatt, S., & Ragiri, P. R. (2021). Security trends in Internet of Things: A survey. SN Applied Sciences,
3(1), 1–14.
Bhattarai, S., & Wang, Y. (2018). End-to-end trust and security for Internet of Things applications.
Computer, 51(4), 20–27. doi:10.1109/MC.2018.2141038
165
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Bhushan, K., & Gupta, B. B. (2019). Distributed denial of service (DDoS) attack mitigation in software
defined network (SDN)-based cloud computing environment. Journal of Ambient Intelligence and Hu-
manized Computing, 10(5), 1985–1997.
Bormann, C., Castellani, A. P., & Shelby, Z. (2012). Coap: An application protocol for billions of tiny
internet nodes. IEEE Internet Computing, 16(2), 62–67.
Brakerski, Z., & Vaikuntanathan, V. (2014). Efficient fully homomorphic encryption from (standard)
LWE. SIAM Journal on Computing, 43(2), 831–871.
Butun, I., Österberg, P., & Song, H. (2019). Security of the Internet of Things: Vulnerabilities, attacks,
and countermeasures. IEEE Communications Surveys and Tutorials, 22(1), 616–644.
Chahid, Y., Benabdellah, M., & Azizi, A. (2017, April). Internet of things security. In 2017 Interna-
tional Conference on Wireless Technologies, Embedded and Intelligent Systems (WITS) (pp. 1-6). IEEE.
doi:10.1109/ICOASE.2018.8548785
Chasaki, D., & Mansour, C. (2015). Security challenges in the internet of things. International Journal
of Space-Based and Situated Computing, 5(3), 141–149.
Chegini, H., & Mahanti, A. (2019, December). A Framework of Automation on Context-Aware Internet
of Things (IoT) Systems. In Proceedings of the 12th IEEE/ACM International Conference on Utility and
Cloud Computing Companion (pp. 157-162). IEEE.
Chen, K., Zhang, S., Li, Z., Zhang, Y., Deng, Q., Ray, S., & Jin, Y. (2018). Internet-of-things security
and vulnerabilities: Taxonomy, challenges, and practice. Journal of Hardware and Systems Security,
2(2), 97–110.
Chen, Y. K. (2012, January). Challenges and opportunities of internet of things. In 17th Asia and South
Pacific design automation conference (pp. 383-388). IEEE.
Chhabra, M., Gupta, B., & Almomani, A. (2013). A novel solution to handle DDOS attack in MANET.
Journal of Information Security, 4(3). DOI: doi:10.4236/jis.2013.43019
Conti, M., Dehghantanha, A., Franke, K., & Watson, S. (2018). Internet of Things security and forensics:
Challenges and opportunities. Academic Press.
Cvitić, I., Peraković, D., Periša, M., & Gupta, B. (2021). Ensemble machine learning approach for classi-
fication of IoT devices in smart home. International Journal of Machine Learning and Cybernetics, 1–24.
Dabbagh, M., & Rayes, A. (2019). Internet of things security and privacy. In Internet of Things from
hype to reality (pp. 211–238). Springer.
Dahiya, A., & Gupta, B. B. (2021). A reputation score policy and Bayesian game theory based incentiv-
ized mechanism for DDoS attacks mitigation and cyber defense. Future Generation Computer Systems,
117, 193–204.
Das, A. K., Zeadally, S., & He, D. (2018). Taxonomy and analysis of security protocols for Internet of
Things. Future Generation Computer Systems, 89, 110–125.
166
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Elhoseny, M., Shankar, K., Lakshmanaprabu, S. K., Maseleno, A., & Arunkumar, N. (2020). Hybrid
optimization with cryptography encryption for medical image security in Internet of Things. Neural
Computing & Applications, 32(15), 10979–10993.
Farooq, M., Waseem, M., Mazhar, S., Khairi, A., & Kamal, T. (2015). A Review on Internet of Things
(IoT). International Journal of Computers and Applications, 113(1), 1–7.
Florea, I., Ruse, L. C., & Rughinis, R. (2017, September). Challenges in security in Internet of Things.
In 2017 16th RoEduNet Conference: Networking in Education and Research (RoEduNet) (pp. 1-5). IEEE.
Gaurav, A., Gupta, B. B., Hsu, C. H., Yamaguchi, S., & Chui, K. T. (2021, January). Fog Layer-based
DDoS attack Detection Approach for Internet-of-Things (IoTs) devices. In 2021 IEEE International
Conference on Consumer Electronics (ICCE) (pp. 1-5). IEEE.
Ge, M., Hong, J. B., Guttmann, W., & Kim, D. S. (2017). A framework for automating security analysis
of the internet of things. Journal of Network and Computer Applications, 83, 12–27.
Ge, M., Hong, J. B., Guttmann, W., & Kim, D. S. (2017). A framework for automating security analysis
of the internet of things. Journal of Network and Computer Applications, 83, 12–27.
Gelenbe, E., Domanska, J., Czàchorski, T., Drosou, A., & Tzovaras, D. (2018, June). Security for internet
of things: The seriot project. In 2018 International Symposium on Networks, Computers and Commu-
nications (ISNCC) (pp. 1-5). IEEE.
Gou, Z., Yamaguchi, S., & Gupta, B. B. (2017). Analysis of various security issues and challenges in
cloud computing environment: a survey. In Identity Theft: Breakthroughs in Research and Practice (pp.
221-247). IGI Global.
Grammatikis, P. I. R., Sarigiannidis, P. G., & Moscholios, I. D. (2019). Securing the Internet of Things:
Challenges, threats and solutions. Internet of Things, 5, 41–70.
Grammatikis, P. I. R., Sarigiannidis, P. G., & Moscholios, I. D. (2019). Securing the Internet of Things:
Challenges, threats and solutions. Internet of Things, 5, 41–70.
Granjal, J., Monteiro, E., & Silva, J. S. (2015). Security for the internet of things: A survey of existing
protocols and open research issues. IEEE Communications Surveys and Tutorials, 17(3), 1294–1312.
Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of things (IoT): A vision, archi-
tectural elements, and future directions. Future Generation Computer Systems, 29(7), 1645–1660.
Gulzar, M., & Abbas, G. (2019, February). Internet of things security: a survey and taxonomy. In 2019
International Conference on Engineering and Emerging Technologies (ICEET) (pp. 1-6). IEEE.
Gupta, B. B., & Quamara, M. (2020). An overview of Internet of Things (IoT): Architectural aspects,
challenges, and protocols. Concurrency and Computation, 32(21), e4946.
Gupta, B. B., & Quamara, M. (2020). An overview of Internet of Things (IoT): Architectural aspects,
challenges, and protocols. Concurrency and Computation, 32(21), e4946.
167
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Gupta, K., & Shukla, S. (2016, February). Internet of Things: Security challenges for next generation
networks. In 2016 International Conference on Innovation and Challenges in Cyber Security (ICICCS-
INBUSH) (pp. 315-318). IEEE.
Hallappanavar, V. L., & Birje, M. N. (2021). A reliable trust computing mechanism in fog computing.
International Journal of Cloud Applications and Computing, 11(1), 1–20.
Hameed, S., Khan, F. I., & Hameed, B. (2019). Understanding security requirements and challenges in
Internet of Things (IoT): A review. Journal of Computer Networks and Communications.
Harbi, Y., Aliouat, Z., Harous, S., Bentaleb, A., & Refoufi, A. (2019). A review of security in internet
of things. Wireless Personal Communications, 108(1), 325–344.
Hassan, W. H. (2019). Current research on Internet of Things (IoT) security: A survey. Computer Net-
works, 148, 283–294. doi:10.1016/j.comnet.2018.11.025
Hassan, W. H. (2019). Current research on Internet of Things (IoT) security: A survey. Computer Net-
works, 148, 283–294.
Hossain, K., Rahman, M., & Roy, S. (2019). Iot data compression and optimization techniques in cloud
storage: Current prospects and future directions. International Journal of Cloud Applications and Com-
puting, 9(2), 43–59.
Hossain, M. M., Fotouhi, M., & Hasan, R. (2015, June). Towards an analysis of security issues, chal-
lenges, and open problems in the internet of things. In 2015 IEEE world congress on services. IEEE.
Hou, J., Qu, L., & Shi, W. (2019). A survey on internet of things security from data perspectives. Com-
puter Networks, 148, 295–306.
Hou, J., Qu, L., & Shi, W. (2019). A survey on internet of things security from data perspectives. Com-
puter Networks, 148, 295–306.
Hu, C., Zhang, J., & Wen, Q. (2011, October). An identity-based personal location system with protected
privacy in IoT. In 2011 4th IEEE International Conference on Broadband Network and Multimedia
Technology (pp. 192-195). IEEE.
Husamuddin, M., & Qayyum, M. (2017, March). Internet of Things: A study on security and privacy
threats. In 2017 2nd International Conference on Anti-Cyber Crimes (ICACC) (pp. 93-97). IEEE.
Jindal, F., Jamar, R., & Churi, P. (2018). Future and challenges of internet of things. International Journal
of Computer Science & Information Technology, 10(2), 13–25.
Jose, D. V., & Vijyalakshmi, A. (2018). An overview of security in Internet of Things. Procedia Com-
puter Science, 143, 744–748. doi:10.1016/j.procs.2018.10.439
Jose, D. V., & Vijyalakshmi, A. (2018). An overview of security in Internet of Things. Procedia Com-
puter Science, 143, 744–748.
168
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Kalyani, G., & Chaudhari, S. (2020). An efficient approach for enhancing security in Internet of Things
using the optimum authentication key. International Journal of Computers and Applications, 42(3),
306–314.
Khattak, H. A., Shah, M. A., Khan, S., Ali, I., & Imran, M. (2019). Perception layer security in Internet
of Things. Future Generation Computer Systems, 100, 144–164. doi:10.1016/j.future.2019.04.038
Khattak, H. A., Shah, M. A., Khan, S., Ali, I., & Imran, M. (2019). Perception layer security in Internet
of Things. Future Generation Computer Systems, 100, 144–164.
Kolias, C., Stavrou, A., Voas, J., Bojanova, I., & Kuhn, R. (2016). Learning Internet-of-Things security”
hands-on. IEEE Security and Privacy, 14(1), 37–46.
Kouicem, D. E., Bouabdallah, A., & Lakhlef, H. (2018). Internet of things security: A top-down survey.
Computer Networks, 141, 199–221.
Kouicem, D. E., Bouabdallah, A., & Lakhlef, H. (2018). Internet of things security: A top-down survey.
Computer Networks, 141, 199–221.
Krichen, M., Cheikhrouhou, O., Lahami, M., Alroobaea, R., & Maâlej, A. J. (2017, November). Towards
a model-based testing framework for the security of internet of things for smart city applications. In
International Conference on Smart Cities, Infrastructure, Technologies and Applications (pp. 360-365).
Springer.
Lee, J. H., & Kim, H. (2017). Security and privacy challenges in the internet of things [security and
privacy matters]. IEEE Consumer Electronics Magazine, 6(3), 134–136.
Li, F., Shi, Y., Shinde, A., Ye, J., & Song, W. (2019). Enhanced cyber-physical security in internet of
things through energy auditing. IEEE Internet of Things Journal, 6(3), 5224–5231.
Li, F., & Xiong, P. (2013). Practical secure communication for integrating wireless sensor networks into
the internet of things. IEEE Sensors Journal, 13(10), 3677–3684.
Li, S., & Da Xu, L. (2017). Securing the internet of things. Syngress.
Li, S., & Da Xu, L. (2017). Securing the internet of things. Syngress.
Li, S., Da Xu, L., & Zhao, S. (2015). The internet of things: A survey. Information Systems Frontiers,
17(2), 243–259.
Li, S., Tryfonas, T., & Li, H. (2016). The Internet of Things: A security point of view. Internet Research.
Li, Z., Yin, X., Geng, Z., Zhang, H., Li, P., Sun, Y., ... Li, L. (2013, January). Research on PKI-like
Protocol for the Internet of Things. In 2013 Fifth International Conference on Measuring Technology
and Mechatronics Automation (pp. 915-918). IEEE.
Lin, J., Yu, W., Zhang, N., Yang, X., Zhang, H., & Zhao, W. (2017). A survey on internet of things:
Architecture, enabling technologies, security and privacy, and applications. IEEE Internet of Things
Journal, 4(5), 1125-1142.
169
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Lin, J., Yu, W., Zhang, N., Yang, X., Zhang, H., & Zhao, W. (2017). A survey on internet of things:
Architecture, enabling technologies, security and privacy, and applications. IEEE Internet of Things
Journal, 4(5), 1125-1142.
Liu, X., Zhao, M., Li, S., Zhang, F., & Trappe, W. (2017). A security framework for the internet of things
in the future internet architecture. Future Internet, 9(3), 27.
Liu, Y., Kuang, Y., Xiao, Y., & Xu, G. (2017). SDN-based data transfer security for Internet of Things.
IEEE Internet of Things Journal, 5(1), 257–268. doi:10.1109/JIOT.2017.2779180
Lv, Z. (2020). Security of internet of things edge devices. Software, Practice & Experience, spe.2806.
doi:10.1002pe.2806
Madakam, S., Lake, V., Lake, V., & Lake, V. (2015). Internet of Things (IoT): A literature review. Journal
of Computer and Communications, 3(05), 164.
Madakam, S., Lake, V., Lake, V., & Lake, V. (2015). Internet of Things (IoT): A literature review. Journal
of Computer and Communications, 3(05), 164.
Madakam, S., Lake, V., Lake, V., & Lake, V. (2015). Internet of Things (IoT): A literature review. Journal
of Computer and Communications, 3(05), 164.
Mahmoud, R., Yousuf, T., Aloul, F., & Zualkernan, I. (2015, December). Internet of things (IoT) security:
Current status, challenges and prospective measures. In 2015 10th International Conference for Internet
Technology and Secured Transactions (ICITST) (pp. 336-341). IEEE.
Makhdoom, I., Abolhasan, M., Lipman, J., Liu, R. P., & Ni, W. (2018). Anatomy of threats to the internet
of things. IEEE Communications Surveys and Tutorials, 21(2), 1636–1675.
Maple, C. (2017). Security and privacy in the internet of things. Journal of Cyber Policy, 2(2), 155–184.
Maple, C. (2017). Security and privacy in the internet of things. Journal of Cyber Policy, 2(2), 155–184.
Mastorakis, G., Mavromoustakis, C. X., & Pallis, E. (2017). Beyond the internet of things (J. M. Batalla,
Ed.). Springer.
Masud, M., Gaba, G. S., Alqahtani, S., Muhammad, G., Gupta, B. B., Kumar, P., & Ghoneim, A. (2020).
A lightweight and robust secure key establishment protocol for internet of medical things in COVID-19
patients care. IEEE Internet of Things Journal.
Mattern, F., & Floerkemeier, C. (2010). From the Internet of Computers to the Internet of Things. In
From active data management to event-based systems and more (pp. 242–259). Springer.
Mavropoulos, O., Mouratidis, H., Fish, A., & Panaousis, E. (2019). Apparatus: A framework for security
analysis in internet of things systems. Ad Hoc Networks, 92, 101743.
McEwen, A., & Cassimally, H. (2013). Designing the internet of things. John Wiley & Sons.
Mendez, D. M., Papapanagiotou, I., & Yang, B. (2017). Internet of things: Survey on security and pri-
vacy. arXiv preprint arXiv:1707.01879.
170
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Mendez Mena, D., Papapanagiotou, I., & Yang, B. (2018). Internet of things: Survey on security. Infor-
mation Security Journal: A Global Perspective, 27(3), 162-182.
Miloslavskaya, N., & Tolstoy, A. (2019). Internet of things: Information security challenges and solu-
tions. Cluster Computing, 22(1), 103–119.
Mirsadeghi, F., Rafsanjani, M. K., & Gupta, B. B. (2020). A trust infrastructure based authentication
method for clustered vehicular ad hoc networks. Peer-to-Peer Networking and Applications, 1–17.
Mishra, A., Gupta, N., & Gupta, B. B. (2021). Defense mechanisms against DDoS attack based on en-
tropy in SDN-cloud using POX controller. Telecommunication Systems, 1–16.
Misra, S., Maheswaran, M., & Hashmi, S. (2017). Security challenges and approaches in internet of
things. Springer International Publishing.
Nandy, T., Idris, M. Y. I. B., Noor, R. M., Kiah, L. M., Lun, L. S., Juma’at, N. B. A., ... Bhattacharyya,
S. (2019). Review on security of Internet of Things authentication mechanism. IEEE Access: Practical
Innovations, Open Solutions, 7, 151054–151089.
Narendrakumar, S., Razaque, A., Patel, V., Almi’ani, M., Rizvi, S. S., & Hans, A. (2018). Token security
for internet of things. International Journal of Embedded Systems, 10(4), 334–343.
Nurse, J. R., Creese, S., & De Roure, D. (2017). Security risk assessment in Internet of Things systems.
IT Professional, 19(5), 20–26.
Obaidat, M. S., Rana, S. P., Maitra, T., Giri, D., & Dutta, S. (2019). Biometric security and internet of
things (IoT). In Biometric-Based Physical and Cybersecurity Systems (pp. 477–509). Springer.
Ogonji, M. M., Okeyo, G., & Wafula, J. M. (2020). A survey on privacy and security of Internet of
Things. Computer Science Review, 38, 100312.
Ongtang, M., McLaughlin, S., Enck, W., & McDaniel, P. (2012). Semantically rich application‐centric
security in Android. Security and Communication Networks, 5(6), 658–673.
Oracevic, A., Dilek, S., & Ozdemir, S. (2017, May). Security in internet of things: A survey. In 2017
International Symposium on Networks, Computers and Communications (ISNCC) (pp. 1-6). IEEE.
10.1109/ISNCC.2017.8072001
Pajouh, H., Dehghantanha, A., Parizi, R. M., Aledhari, M., & Karimipour, H. (2019). A survey on internet
of things security: Requirements, challenges, and solutions. Internet of Things, 100129.
Patwary, A. A. N., Fu, A., Battula, S. K., Naha, R. K., Garg, S., & Mahanti, A. (2020). FogAuthChain:
A secure location-based authentication scheme in fog computing environments using Blockchain. Com-
puter Communications, 162, 212–224.
Ray, P. P. (2018). A survey on Internet of Things architectures. Journal of King Saud University-Computer
and Information Sciences, 30(3), 291–319.
Razzaq, M. A., Gill, S. H., Qureshi, M. A., & Ullah, S. (2017). Security issues in the Internet of Things
(IoT): A comprehensive study. International Journal of Advanced Computer Science and Applications,
8(6), 383.
171
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Razzaq, M. A., Gill, S. H., Qureshi, M. A., & Ullah, S. (2017). Security issues in the Internet of Things
(IoT): A comprehensive study. International Journal of Advanced Computer Science and Applications,
8(6), 383.
Ren, Z., Liu, X., Ye, R., & Zhang, T. (2017, July). Security and privacy on internet of things. In 2017 7th
IEEE International Conference on Electronics Information and Emergency Communication (ICEIEC)
(pp. 140-144). IEEE.
Rizvi, S., Kurtz, A., Pfeffer, J., & Rizvi, M. (2018, August). Securing the internet of things (IoT): A
security taxonomy for IoT. In 2018 17th IEEE International Conference On Trust, Security And Privacy
In Computing And Communications/12th IEEE International Conference On Big Data Science And
Engineering (TrustCom/BigDataSE) (pp. 163-168). IEEE.
Robertazzi, T. G. (2017). Software-defined networking. In Introduction to Computer Networking (pp.
81–87). Springer.
Roman, R., Najera, P., & Lopez, J. (2011). Securing the internet of things. Computer, 44(9), 51–58.
Sadeeq, M. A., Zeebaree, S. R., Qashi, R., Ahmed, S. H., & Jacksi, K. (2018, October). Internet of Things
security: a survey. In 2018 International Conference on Advanced Science and Engineering (ICOASE)
(pp. 162-166). IEEE.
Sadeghi, A. R., Wachsmann, C., & Waidner, M. (2015, June). Security and privacy challenges in industrial
internet of things. In 2015 52nd ACM/EDAC/IEEE Design Automation Conference (DAC) (pp. 1-6). IEEE.
Saha, H. N., Mandal, A., & Sinha, A. (2017, January). Recent trends in the Internet of Things. In 2017
IEEE 7th annual computing and communication workshop and conference (CCWC) (pp. 1-4). IEEE.
Sahmim, S., & Gharsellaoui, H. (2017). Privacy and security in internet-based computing: cloud com-
puting, internet of things, cloud of things: a review. Procedia Computer Science, 112, 1516–1522.
doi:10.1016/j.procs.2017.08.050
Sain, M., Kang, Y. J., & Lee, H. J. (2017, February). Survey on security in Internet of Things: State of
the art and challenges. In 2017 19th International conference on advanced communication technology
(ICACT) (pp. 699-704). IEEE.
Salam, A. (2020). Internet of things in agricultural innovation and security. In Internet of Things for
Sustainable Community Development (pp. 71–112). Springer.
Salhi, D. E., Tari, A., & Kechadi, M. T. (2021). Using Clustering for Forensics Analysis on Internet of
Things. International Journal of Software Science and Computational Intelligence, 13(1), 56–71.
Salman, T., & Jain, R. (2019). A survey of protocols and standards for internet of things. arXiv preprint
arXiv:1903.11549.
Samaila, M. G., Neto, M., Fernandes, D. A., Freire, M. M., & Inácio, P. R. (2017). Security challenges
of the Internet of Things. In Beyond the Internet of Things (pp. 53–82). Springer.
Samaila, M. G., Neto, M., Fernandes, D. A., Freire, M. M., & Inácio, P. R. (2017). Security challenges
of the Internet of Things. In Beyond the Internet of Things (pp. 53–82). Springer.
172
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Sarma, A., Matos, A., Girao, J., & Aguiar, R. L. (2008). Virtual identity framework for telecom infra-
structures. Wireless Personal Communications, 45(4), 521–543.
Sfar, A. R., Natalizio, E., Challal, Y., & Chtourou, Z. (2018). A roadmap for security challenges in the
Internet of Things. Digital Communications and Networks, 4(2), 118–137. doi:10.1016/j.dcan.2017.04.003
Sfar, A. R., Natalizio, E., Challal, Y., & Chtourou, Z. (2018). A roadmap for security challenges in the
Internet of Things. Digital Communications and Networks, 4(2), 118–137.
Sha, K., Wei, W., Yang, T. A., Wang, Z., & Shi, W. (2018). On security challenges and open issues in
Internet of Things. Future Generation Computer Systems, 83, 326–337.
Sial, A., Singh, A., & Mahanti, A. (2019). Detecting anomalous energy consumption using contextual
analysis of smart meter data. Wireless Networks, 1–18.
Sial, A., Singh, A., Mahanti, A., & Gong, M. (2018, April). Heuristics-Based Detection of Abnormal
Energy Consumption. In International Conference on Smart Grid Inspired Future Technologies (pp.
21-31). Springer.
Sicari, S., Rizzardi, A., Grieco, L. A., & Coen-Porisini, A. (2015). Security, privacy and trust in Internet
of Things: The road ahead. Computer Networks, 76, 146–164.
Singh, N., & Vardhan, M. (2019). Distributed ledger technology based property transaction system
with support for iot devices. International Journal of Cloud Applications and Computing, 9(2), 60–78.
Sklavos, N., & Zaharakis, I. D. (2016, November). Cryptography and security in internet of things (iots):
Models, schemes, and implementations. In 2016 8th IFIP International Conference on New Technolo-
gies, Mobility and Security (NTMS) (pp. 1-2). IEEE.
Stankovic, J. A. (2014). Research directions for the internet of things. IEEE Internet of Things Journal,
1(1), 3–9.
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based big data secure management in the fog
over a 6G wireless network. IEEE Internet of Things Journal, 8(7), 5164–5171.
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based Big Data secure management in the
Fog over a 6G Wireless Network. IEEE Internet of Things Journal.
Sun, W., Cai, Z., Li, Y., Liu, F., Fang, S., & Wang, G. (2018). Security and privacy in the medical internet
of things: A review. Security and Communication Networks.
Sun, W., Cai, Z., Li, Y., Liu, F., Fang, S., & Wang, G. (2018). Security and privacy in the medical internet
of things: A review. Security and Communication Networks.
Sundaram, B. V., Ramnath, M., Prasanth, M., & Sundaram, V. (2015, March). Encryption and hash
based security in Internet of Things. In 2015 3rd International Conference on Signal Processing, Com-
munication and Networking (ICSCN) (pp. 1-6). IEEE.
Suto, L. (2010). Analyzing the accuracy and time costs of web application security scanners. Academic
Press.
173
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Tahsien, S. M., Karimipour, H., & Spachos, P. (2020). Machine learning based solutions for security
of Internet of Things (IoT): A survey. Journal of Network and Computer Applications, 161, 102630.
Tankard, C. (2015). The security issues of the Internet of Things. Computer Fraud & Security, 2015(9),
11–14.
Tewari, A., & Gupta, B. B. (2017). A lightweight mutual authentication protocol based on elliptic curve
cryptography for IoT devices. International Journal of Advanced Intelligence Paradigms, 9(2-3), 111–121.
Tewari, A., & Gupta, B. B. (2020). Security, privacy and trust of different layers in Internet-of-Things
(IoTs) framework. Future Generation Computer Systems, 108, 909–920.
Tewari, A., & Gupta, B. B. (2020). Security, privacy and trust of different layers in Internet-of-Things
(IoTs) framework. Future Generation Computer Systems, 108, 909–920.
Tzafestas, S. G. (2018). Ethics and law in the internet of things world. Smart Cities, 1(1), 98-120.
Van Oorschot, P. C., & Smith, S. W. (2019). The internet of things: Security challenges. IEEE Security
and Privacy, 17(5), 7–9.
Vashi, S., Ram, J., Modi, J., Verma, S., & Prakash, C. (2017, February). Internet of Things (IoT): A
vision, architectural elements, and security issues. In 2017 international conference on I-SMAC (IoT in
Social, Mobile, Analytics and Cloud)(I-SMAC) (pp. 492-496). IEEE.
Vasilomanolakis, E., Daubert, J., Luthra, M., Gazis, V., Wiesmaier, A., & Kikiras, P. (2015, Septem-
ber). On the security and privacy of Internet of Things architectures and systems. In 2015 International
Workshop on Secure Internet of Things (SIoT) (pp. 49-57). IEEE.
Virat, M. S., Bindu, S. M., Aishwarya, B., Dhanush, B. N., & Kounte, M. R. (2018, May). Security and
privacy challenges in internet of things. In 2018 2nd International Conference on Trends in Electronics
and Informatics (ICOEI) (pp. 454-460). IEEE.
Wang, N., Jiang, T., Li, W., & Lv, S. (2017). Physical-layer security in Internet of Things based on
compressed sensing and frequency selection. IET Communications, 11(9), 1431–1437.
Weber, R. H. (2010). Internet of Things–New security and privacy challenges. Computer Law & Security
Review, 26(1), 23–30.
Wei, W., Yang, A. T., Shi, W., & Sha, K. (2016, October). Security in internet of things: Opportunities
and challenges. In 2016 International Conference on Identification, Information and Knowledge in the
Internet of Things (IIKI) (pp. 512-518). IEEE.
Wortmann, F., & Flüchter, K. (2015). Internet of things. Business & Information Systems Engineering,
57(3), 221–224.
Yang, Y., Wu, L., Yin, G., Li, L., & Zhao, H. (2017). A survey on security and privacy issues in Internet-
of-Things. IEEE Internet of Things Journal, 4(5), 1250–1258.
174
A Survey on Emerging Security Issues, Challenges, and Solutions for Internet of Things (IoTs)
Yaqoob, I., & Ahmed, E., ur Rehman, M. H., Ahmed, A. I. A., Al-garadi, M. A., Imran, M., & Guizani,
M. (2017). The rise of ransomware and emerging security challenges in the Internet of Things. Computer
Networks, 129, 444–458.
Yousefi, A., & Jameii, S. M. (2017, May). Improving the security of internet of things using encryption
algorithms. In 2017 International Conference on IoT and Application (ICIOT) (pp. 1-5). IEEE. 10.1109/
ICIOTA.2017.8073627
Yousefi, A., & Jameii, S. M. (2017, May). Improving the security of internet of things using encryption
algorithms. In 2017 International Conference on IoT and Application (ICIOT) (pp. 1-5). IEEE.
Yousuf, O., & Mir, R. N. (2019). A survey on the internet of things security: State-of-art, architecture,
issues and countermeasures. Information & Computer Security.
Yu, Y., Li, Y., Tian, J., & Liu, J. (2018). Blockchain-based solutions to security and privacy issues in
the internet of things. IEEE Wireless Communications, 25(6), 12–18.
175
176
Chapter 10
SecBrain:
A Framework to Detect Cyberattacks
Revealing Sensitive Data in
Brain-Computer Interfaces
ABSTRACT
In recent years, the growth of brain-computer interfaces (BCIs) has been remarkable in specific appli-
cation fields, such as the medical sector or the entertainment industry. Most of these fields use evoked
potentials, like P300, to obtain neural data able to handle prostheses or achieve greater immersion
experience in videogames. The natural use of BCI involves the management of sensitive users’ infor-
mation as behaviors, emotions, or thoughts. In this context, new security breaches in BCI are offering
cybercriminals the possibility of collecting sensitive data and affecting subjects’ physical integrity, which
are critical issues. For all these reasons, the fact of applying efficient cybersecurity mechanisms has
become a main challenge. To improve this challenge, this chapter proposes a framework able to detect
cyberattacks affecting one of the most typical scenarios of BCI, the generation of P300 through visual
stimuli. A pool of experiments demonstrates the performance of the proposed framework.
DOI: 10.4018/978-1-7998-7789-9.ch010
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
SecBrain
INTRODUCTION
Brain-Computer Interfaces (BCIs) are devices that enable two-way communication between an individual’s
brain and external devices. This bidirectional connection allows two different functionalities in terms of
BCI usage. The first one is focused on the acquisition of neuronal activity produced by an individual and
its transmission to a computer for analysis and processing. The second is given by the stimulation and
inhibition of brain activity to regulate abnormal impulses or improve motor actions at a neuronal level.
Traditionally, the usage of BCI has been aligned with the medical field. With this technology, many
advances have been made in neuropsychology and neurophysiology. BCI has contributed to the treat-
ment of neurodegenerative diseases by analyzing the brain state, such as epilepsy and the autonomic
nervous system (Liberati et al., 2012; Simon et al., 2011). Over the years, BCI technology has undergone
significant technological evolution. Thanks to numerous studies, BCI has increased its application fields
and has started to be used in other scenarios than medicine. One of these scenarios is the entertainment
and video game industry (Ahn et al., 2014; Finke et al., 2009). Another sector exploring the use of BCI
is the military one, where studies are aiming to allow the telepathic handling of multiple drones at a
distance (Al-Nuaimi et al., 2020) or even exoskeletons (Crea et al., 2018).
Most of the scenarios functionality is based on capturing and processing the electroencephalography
(EEG) signal and evoked potentials. Event-related potentials (ERPs) are signal patterns automatically
generated by the brain when stimuli are presented to the person. Different types of potentials depend on
the trigger action performed: visual, auditory, somatosensory, or cognitive. The study of these potentials
has made it possible to obtain information about the subject, such as his/her emotional state, neurological
problems, dependencies, or even private information.
One of the most well-known and used ERPs in brain recording is P300 (or P3). P300 is related to the
visualization of stimuli known by the person. It is produced between 250-500 ms after the visualization
of each known-stimulus and has a positive signal peak. One of the most common ways of provoking
this potential is through the Oddball paradigm. The Oddball paradigm shows a series of known stimuli
belonging to a more extensive set of unknown stimuli. At this point, it is important to mention that the
captured EEG and the labeling of the P300 are susceptible for the user. This problem is aggravated due
to the lack of frameworks that consider security aspects such as authentication, confidentiality, and data
integrity. In this context, attackers could turn their attention to the BCIs to carry out malicious actions.
This work is motivated by the limitations of current frameworks, which do not provide security mecha-
nisms to ensure the integrity of transmitted data or users’ privacy (Ghoneim et al., 2018). Many times,
this leads to a malfunction of the actions carried out by the BCI or to leak sensitive information of the
individual. Similarly, current EEG-based BCI frameworks do not provide authentication mechanisms,
so an attacker could impersonate the legitimate user to adapt the BCI functionality with malicious data.
Besides, there is no standard or specific protocol for the secure development of BCI applications, causing
a significant weakness in the software and its interaction with the hardware in many current alternatives.
177
SecBrain
Contribution
To improve the previous challenges, this work designs and implements an intelligent and automatic
framework, called from now SecBrain, that manages each phase of the BCI life cycle (communication
and acquisition of EEG, data pre-processing, and P300 detection) to improve the limitations of existing
BCI frameworks. SecBrain also considers the cybersecurity of the BCI life cycle, implementing security
mechanisms to ensure the proper and secure framework functioning. The objective is to ensure data
integrity and confidentiality throughout each phase of the BCI cycle and the secure storage of them
afterward. For this purpose, the functionality is divided into several security layers. On the one hand,
SecBrain encapsulates the EEG signal flows acquired by the BCI, ensuring the data confidentiality and
integrity directly from the subject. This process is relevant since EEG travels from the headset equipped
with electrodes to an external device where the data are acquired, processed, and stored, causing an at-
tacker to manipulate the transmitted data or deliberately introduce new ones. On the other hand, SecBrain
provides a secure scenario for BCI applications where visual stimuli predominate. The framework logic
allows adapting the system in a personalized way according to the analysis purpose. It also guarantees
the synchronization of stimuli with the EEG signal previously acquired by the BCI, including security
mechanisms/capabilities. In this way, the aim is to avoid the materialization of malware in certain phases
of the BCI cycle, affecting the acquired brain reading or denying the service in certain moments of the
experiment. In order to measure the framework efficiency and effectiveness, this chapter proposes a
scenario where an attacker performs different malware-based attacks on the proposed structure. The
attacks contemplated are designed to affect the integrity of the data transmitted. In contrast, others are
intended to infer the proper functioning of the BCI or manage the communication and interaction for
the attacker’s deliberation. These attack scenarios are applied to a specific use case, generating P300
potentials through the Oddball paradigm. This scenario is among the most common BCI applications
as it provides a high level of neuronal information. This allows contrasting the security provided by
SecBrain and the automatic detection measures.
The remainder of this chapter is structured as follows. The Background section presents the current
state of the art regarding BCI, its benefits, and current cybersecurity issues. After that, the SecBrain
Framework: Design and Deployments section introduces the design and implementation details of the
proposed framework. In addition, a real experiment using SecBrain implementation is included. Section
Future Research Directions highlights the current trend and future evolution of BCI, indicating possible
opportunities in this area. Finally, the Conclusion section presents a summary of the chapter.
Background
Brain-Computer Interface (BCI) is a technology that establishes two-way communication between the
brain and an external device. One of its applications is to capture the biometric signal produced by a
subject’s brain (Moreno et al., 2019). This signal is then interpreted and processed to give meaning to
the information acquired. The main idea of this scenario is to analyze the signal produced and obtain
its essential characteristics to finally transform it into a command or order that a computer can process
and execute. Therefore, in many areas, such as medicine, it is considered a fundamental tool for treating
patients.
Currently, other sectors, such as the leisure and entertainment sector, are taking great interest in us-
ing this technology. The interaction, the immersion provided, and, in short, the gameplay experience
178
SecBrain
are fundamental pillars that BCI tries to improve by adapting its cycle to the purpose of the video game.
Finkie et al. (2009) translated the brain signals into movements of a character within a video game. Alba
et al. (2011) applied something like the previous one but in the mythical game “Pong” style. In recent
years the use of BCI has been complemented with virtual reality. The main advantage is to improve
the immersive visual experience of the subject and, consequently, enhancing the data obtained. In this
sense, Fu et al. (2020) propose a cultural experience scenario, creating a room escape game using the
EEG signal to interact virtually.
In cybersecurity, the use of this technology is also growing, intending to create new authentication
mechanisms. Curran et al. (2016) used a commercial BCI, EarEEG, placed in the ear that allows for
authentication based on tasks such as breathing with eyes closed, thinking of a song, or imagining a face.
Rathi et al. (2021) proposed a robust authentication system based on EEG signals and the generation of
the P300 potential during the visualization of different images, breaking with the traditional password
and token schemes. The EEG signal generation is unique per subject making it a reliable and suitable
system to protect assets. Finally, well-known cybersecurity attacks from network communications could
be applied to BCI scenarios (Gupta et al., 2020).
BCI Lifecycle
The BCI maintains direct communication between the brain and a computer, opening a range of pos-
sibilities for interaction. Recording the brain signals using electrical activity is the most widely used
technique in non-invasive systems. The electroencephalography technique or EEG is an excellent and
inexpensive alternative (Badcock et al., 2013; Aydemir et al., 2015), with practically no risks and quite
precise results (Millan et al., 2007; Polich et al., 1985). All this electrical activity produced will be
transmitted to the computer, which will process and finally translate into a control command. A control
command is a term given to a particular instruction that has been programmed or implemented in a
computer. It carries out a functionality, for example, moving a prosthesis or a character within a video
game. The literature gathers different proposals on structuring the phases or layers of the BCI life cycle.
In this sense, López Bernal et al. (2021) proposed a bidirectional scheme, on the one hand, the neural
data acquisition process and, on the other, brain stimulation. Both processes go through four phases:
EEG signal generation, acquisition, processing, and end-user application.
Evoked Potentials
It is necessary to pay attention to the evoked potentials (Chiappa & Ropper, 1982, p. 1147) for reading
brain signals. Event-related potentials (ERPs) are brain responses to sensory stimuli (visual, auditory,
motor, or somesthetic) generated involuntarily by the brain. The normality or otherwise of these re-
sponses expresses the functional state of the visual, acoustic, motor, or somesthetic pathway examined.
Since the signal is tiny (microvolts), it is necessary to give many stimuli to average the responses. These
responses can reveal the cause of the discomfort suffered by a subject, which is why it has been con-
sidered necessary to carry out evoked potentials in their different modalities. In Alzheimer’s disease,
evoked potentials have been studied to determine the disease in patients. Na et al. (2020) have identified
deficits in visual attention using pre-attentive ERPs in patients with the disease relative to a group of
cognitively normal adults.
179
SecBrain
One of the evoked potentials is the P300. The P300 wave is a positive deviation or increase in voltage
directly related to the appearance of an event that provokes a subject’s stimulus. This potential was first
referred to as a long latency potential called the “text positive component”. Later, Polich et al. (1996)
defined it as P300 due to the 300ms latency in which the increase in voltage appeared in young subjects
without psychic problems after the stimulation.
The Oddball paradigm is the technique that is usually used to obtain the P300 potential (Jang et al.,
2011; Eligiusz Wronka et al., 2008). This procedure, located in the category of visual events, consists
of showing a string of images to the subject with a frequency of approximately 250-500ms. About 10-
20% are known or familiar.
This technique is based on a continuous sample of stimuli that could be cataloged as usual (Non-Target).
Among these, one or more uncommon stimuli (Target) will appear (Donchin et al., 1979). The Oddball
paradigm is based on the presentation of stimuli that are discordant to what is usually presented. These
stimuli can be both sound (Cass et al., 1997) and visual (Choo & May, 2014). This procedure allows
detecting when a trigger is recognized by the subject, creating a characteristic wave, the P300, instead
of neutral waves generated by unfamiliar stimuli. With this premise, applications can be made that range
from authentication by known images, spellers, or even to detect people with attention deficiencies.
Evoked potentials have been well studied to provide solutions to various problems. This research will
be shown below, explaining what the P300 is about and how it is used. One of these investigations is
that of “authentication”, where it is intended to use the P300 potential for subject authentication using
cognitive biometrics. It uses a series of letters, images, or colors, which only this subject knows. This
work aims to protect, together with other methods, applications that need a high-security level.
Something similar aims to “authenticate faces”, developing a system based on showing a series of
human faces, both unknown and known, and studying the subject’s response to them. In this way, it is
possible to identify the individual who is trying to authenticate himself. Other interesting articles on
this subject are Gupta et al. (2012) and Kaongoen et al. (2020). Another of the most widespread uses
given to the evoked potential P300 is to detect neurodegenerative diseases by studying the latency and
amplitude of the P300 wave. Some of the disorders can be Alzheimer’s, schizophrenia, or Parkinson’s.
Finally, some of the more varied uses that have been made of the potential evoked by P300 are sub-
sequently commented. For example, Nurseitov et al. (2017) proposed the control of a robot, based on
P300, which gave the possibility of ordering the robot to move in any of the four directions (left, right,
forward, or backward). This can be extrapolated to the improvement of the quality of life of people with
locomotion problems or reduced mobility, offering the possibility of moving a wheelchair (Iturrate et
al., 2019) if necessary, or the opportunity to control a robot to assist these people as reduced mobility
Arrichiello et al. (2017).
BCI Frameworks
This section compares the existing BCI frameworks, giving an overview of the characteristics of each
solution (Table 1). The two main aspects that have been considered are (1) implementing one or more
communications for the reception of data from the BCI platform and (2) the inclusion of a wide variety
of filtering and signal processing applications in a personalized and straightforward way.
180
SecBrain
OpenBCI GUI (The OpenBCI GUI, 2020) is compatible with OpenBCI hardware. It implements
substantial improvements in the final data representation, and it is compatible with LSL communications.
Additionally, it provides the program source code and the ability to create custom modules that integrate
with the desktop application. OpenViBe (OpenViBE, 2020) is the best alternative to launch pre-set experi-
ments related to the Oddball paradigm. It offers schemes and customizable modules. Python MNE is the
most used alternative in EEG signal processing, offering a high degree of modification. It also allows the
use of complementary Python libraries to treat the data, apply filtering functions or noise applications
to the signal for further study. Letswave (Letswave7, 2018), EEGLab (EEGLAB, 2020), and BCILab
(BCI Lab, 2020) are complements to the Matlab (MATLAB, 2020) application. They are the most used
alternatives for treating the data with Matlab, offering compatibility of import and export of the data.
The most remarkable feature of these options is the pre-configuration already done of most of the filters.
Cybersecurity on BCI
Any device interconnected with another or towards the Internet is subject to possible attacks questioning
cybersecurity. Several cyberattacks appear to extract sensitive information such as passwords, religious
beliefs, political orientations, phone numbers, or credit card numbers (Lopez Bernal et al., 2020). Ac-
cording to the stimulus presented, most of these attacks are based on provoking a specific response in
the brain. In this sense, Martinovic et al. (2012) carried out some experiments to steal critical subject
information such as the bank 4-digit PIN code, data, and even the person’s place of residence, in all of
them using the generation of the P300 potential through visual stimuli.
The literature has not dealt in depth with the materialization of this kind of attacks, leading to insuf-
ficient protection and detection of threats. Meng et al. (2019) leave aside obtaining personal information
from the subject to focus on the integrity of the captured data, applying deliberate modifications to alter
them. Similarly, Sundararajan (2018) developed a laboratory scenario that executed tests with different
attacks. These attacks were: 1) passive listening, which intercepted the data without the user being aware,
2) active interception, where it collected the data and could discard or resend it, 3) denial of service and
4) data modification, where it ordered the data, modified it and resent it to obtain a different response.
As can be seen, all of them addressed possible attacks on these devices to steal information or cor-
rupt the interaction. The implementation of security mechanisms for BCI devices is scarcely addressed
theoretically in the literature. In this context, Ajrawi et al. (2021) proposed a theoretical framework
based on RFID to identify the subject’s brain activity in real-time. The study focuses on the medical
sector, deploying a scenario of EEG acquisition in patients. In short, it defines a theoretical approach that
could improve and provide secure BCI applications. However, there is an absence of works presenting
a practical implementation of active or proactive attack detection mechanisms and considering aspects
such as the integrity of the transmitted data or the subject’s privacy. In this sense, recent studies have
reiterated the lack of cybersecurity in BCI devices and developed applications. Chaudhary et al. (2018)
detailed the ethical and safety issues of commercial BCIs such as BioSemi Active or Emotiv EPOC. In
particular, the authors analyzed the BCI development process and possible security measures needed to
mitigate threats to end-user privacy.
181
SecBrain
The main objective of this section is to explain the design and implementation details of SecBrain, our
framework able to capture the EEG signal through the BCI, pre-process it, and detect the P300. A use
case is also presented to demonstrate the added-value and feasibility of SecBrain. Finally, this section
also carries out adverse attacks on the EEG to examine its implications and the possibility of detecting
them by SecBrain at any stage.
182
SecBrain
Use Case
The proposed use case presents a subject watching visual stimuli that are part of a video. As seen
throughout this section, SecBrain monitors, acquires, and analyzes the EEG when the subject watches
the video (containing known and unknown stimuli). Finally, through the P300, SecBrain automatically
detects if each presented stimulus is known (Target) or not (Non-Target). For the creation of the video,
the Oddball paradigm has been implemented, and two types of images are shown:
• Target image: it is shown at the beginning of the experiment and is the known stimulus. It will be
repeated during the video, with a probability indicated at the beginning.
• Non-Target images: rest of the pictures shown in the video and represent the unknown stimuli.
They generate the Non-Target signal in the capture of the EEG signal. This signal differs from that
produced by the potential evoked by a viewing of the Target image.
The Non-Target images used in the creation of the video are random and meaningless for the subject.
They belong to the categories of landscapes, animals, or unknown subjects. One of them is obtained
as a target image and is shown at the beginning of the video. The Target image is the only one that is
repeated several times randomly during the display.
In the previous scenario setup, the authors have performed three different experiments. In each ex-
periment, wholly random and additional images are added, and another visual stimulus organization is
established. SecBrain obtains random images, with the following implementation has been carried out
with two different modes:
• Manual mode (Mode 1 onwards) allows the possibility of including the images that the framework
administrator deems appropriate. This mode considers the dimensions of the monitor where the
experiment so that they all have the same aspect ratio.
• Automatic mode (Mode 2 onwards). It allows obtaining different images to include in the experi-
ment. This mode downloads some random images indicated as a parameter in the framework, one
of which will be selected as Target, while the others will be included as Non-Target
For the implementation of Mode 2, Unsplash API (Unsplash, 2020) has been used to obtain random
images of specific dimensions. All the photos have a random order along with the display. The propor-
tion given to the type of target image is the determining factor for its appearance in the experiment de-
veloped. A ratio of 0.1 in an experiment with ten images implies one Target image and nine Non-Target
images. To conclude, it is important to mention that the previous configuration setup is flexible and can
be modified and configured at the beginning of the experiment.
Framework
This section details the design and implementation of SecBrain, whose objective is to create a secure
and stable environment to develop scenarios based on the detection of P300 generated through visual
stimuli. For that, the following phases of the BCI cycle (EEG acquisition, signal processing, and the
identification of the P300) and the detection of possible threats during the process are considered. The
conceptual design of the framework is shown in Figure 1.
183
SecBrain
The use of the Oddball paradigm in the experiments implies the recording of how each stimulus is
shown to the subject and what type of stimulus has been shown (Target or Non-Target) to make later a
detailed study of the generation of the P300. This section deals with the details of signal acquisition and
its corresponding synchronization.
The EEG signal acquisition is performed through electrodes placed on the scalp thanks to a non-
invasive BCI. This work has used the OpenBCI Headset, whose communication is performed via an
external module provided by the OpenBCI software. The primary function of this module is to derive
the data stream from the BCI controller and obtained by the USB Dongle (Figure 2) to an LSL commu-
nication. This adaptation allows the independence of the framework from the hardware used, with the
possibility of using other BCIs different from the one used in this chapter. The framework controls each
data flow originated by the BCI so that the acquisition phase will be closely related to the framework
implementation. The framework generates an independent thread that remains listening to any EEG data
stream for the reception of the EEG signal via LSL.
The Python library PyLSL is handy for this part of the acquisition. When it detects incoming data
(Inlet), it receives it and stores it in a structured and organized way. The structure of Table 2 allows it to
be correctly matched with the information received, so the acquisition of data for subsequent analysis is
straightforward. The first two columns indicate time marks for each acquired sample, the first relative to
the beginning of the experiment and the second one absolute (both in seconds). The following columns
represent the value obtained per channel at each instant of time. The last column represents the Stream
Stimulus, which maintains the synchronization of visual stimuli concerning each sample of the data,
being the value 0 by default, 1 for a Target image, and 2 for a Non-Target image. The structure creation
has been based on the definition of the ASCII standard in the representation of EEG data, the same
format as when exporting raw data, stored in a .txt file, from the OpenBCI GUI. When the experiment
starts, the framework proceeds to the synchronization of the visual elements shown on the screen with
the EEG signal received by the thread through LSL. The signal-stimulus synchronization is done at run
time, saving the synchronized data in internal structures.
184
SecBrain
Figure 2. OpenBCI Cyton hardware: left, different elements of the Cyton hardware; right, placement of
the eight channels for acquisition
Table 2. Extract with the defined structure to store the EEG capture data
Stimuli
Time Timestamp FP1 … O2
type
82.435 1594894785.1193922 14415.1533203125 … 17168.91015625 0
82.439 1594894785.1193976 14379.2568359375 … 17215.15625 1
82.447 1594894785.5996578 14589.876953125 … 17114.439453125 0
SecBrain uses the epoch elements to synchronize the signal with each of the events produced during
the experiment. Therefore, an epoch is a data structure that stores in a standard and synchronized way
the following aspects: 1) EEG signal of each channel acquired by the computer and 2) events produced
during the experiment display. The main advantage it offers is the representation of specific sections of
the signal. Each segment will contain the event produced for further analysis. The framework has been
initialized to obtain portions of the EEG signal that take as start 0.1 seconds before the event occurs and
0.8 seconds after (Figure 4). Also, it incorporates rejection parameters based on peak-to-peak amplitude.
By default, a rejection parameter with a value of 100e-6 is applied to each electrode that intervenes at
each time (Figure 5). A representation scheme is created to assign a value of 1 for Target and 2 for Non-
Target; the value 0 will be the default for each one of the samples with no relation to the events. This
procedure allows the different events produced to be differentiated and studied individually or compared
between them. Figure 4 shows the first 20 of the 96 total periods stored by the framework in one of the
acquisitions made. This type of independent representation will also help determine if the filters and
treatments applied over the signal have reduced noise.
The attack scenario presented in Figure 3 focuses on the conceptual design of the framework. The
figure details those areas that are vulnerable due to the use of malware. The first of these can be mate-
rialized in the communication between the BCI headset and the receiver placed in an external device.
The second one is between the receiving device and the framework, where the external device infection
185
SecBrain
Figure 3. The proposed architecture and its possible attack regions in the acquired data—deployment
of threat detection systems
could cause severe data integrity and confidentiality problems. The third of these can occur during the
process of synchronization and storage of the data.
The signal acquisition component is one of the main functionalities of the infrastructure. Therefore, it
is compulsory to guarantee a minimum level of security on the device and during communication. Two
different types of threats can occur. The first one materializes when the attackers know everything about
the BCI application and know how they can affect its functionality. The second of these occurs when
the attacker uses generic malware that attempts to spoil existing communication without considering
the nature of the data transmitted and stored.
Regarding the first perspective, a more comprehensive range of possibilities is opened. In this sense,
an attacker can deploy a system like the one provided by the BCI and generate an intermediate entity
that negotiates the different BCI parameters. The BCI device would transmit information directly to
an entity that is not legitimate, the definition of Man-in-the-Middle attacks. The receiving computer is
infected with malware that passes itself off as legitimate software for receiving the data. This malware
will capture all the data that the BCI sends, and along with the flow of images, the attacker could link
the P300 provoked and obtain sensitive information from the subject. The modification of the data can
be a triggering event in both scenarios. However, only in the first perspective, the attacker will enter
the data needed to compromise the system to gain total control. The total system hijacking occurs when
the attacker knows the process to be followed and what exactly would be the parameter that disturbs the
proper functioning. Careful preparation of malware for the P300 scenario described above could lead to
the generation of false P300s or the attenuation of these. In the same way that the data can be modified,
186
SecBrain
Figure 4. First twenty times of the EEG tagged signal: “1” for Target event and “2” for non-target
187
SecBrain
specific changes could be applied to the images shown, and the stimuli provoked in the subjects studied.
The purpose of this could be to obtain compromised information.
In terms of the second scenario, the attacker has no information about how a particular BCI system
works and its application. In this circumstance, the attackers are more limited in their ability to break into
the system. However, malicious malware could start making deliberate modifications (without knowing
the nature of the data). The main objective is to cause a malfunction of the legitimate application or
even a denial of service (DoS). It is, therefore, necessary to implement a series of countermeasures to
mitigate or avoid these issues. First, the framework must guarantee that the set of images introduced in
the experiment has not undergone any modification. The images are the main asset of the experiment
using the Oddball paradigm technique. All these images are loaded into container structures internally
in the framework. A hash is applied to each image to avoid any kind of deliberate modification (adding
images, removing pictures, or modifying image parameters). It is necessary to ensure the integrity of
the structure in transmission and storage.
Based on these requirements, PI= I1,I2,I3,I4,…,NN represents the internal structure of images. Ad-
ditionally, a SHA-256 hash function noted as Hash(), is applied over the PI structure, as follows:
EI 1 Hash I n
n
Figure 6 shows graphically the flow and processes followed by the data. Initially, the acquisition
module generates the hash of the set of images. In the same way, when showing the pictures on the
screen, the hash will be generated. If both coincide, the data will be acquired and saved; otherwise, an
alert will be generated.
With this approach, the framework guarantees the integrity of the data displayed and, consequently,
warns of any modification in the data to alert the subject. This procedure increases security in the use
of BCI applications, as well as warns of the presence of any threat.
188
SecBrain
As a second procedure, the confidentiality of transmitted and received data is guaranteed. To this
end, cryptographic processes are incorporated at the socket level. The framework gets information
through the LSL protocol (Section Signal acquisition component). This protocol does not include se-
curity mechanisms, so it is convenient to treat the received data in such a way as to guarantee the clear
reading of the data during the development of the experiment or in later phases of the BCI cycle, once
the data have been stored.
The mechanism to be used is based on encrypting the data flows on reception. When the BCI obtains
the voltage from each of the electrodes located on the scalp, the data collected from each electrode are
sent to an external device that acquires the data. Sometimes the external device is simply an intermediate
device that allows reaching the final device. The data encryption must be taken into account to avoid
malware that allows sniffing of the data from contact with the first device.
In particular, the symmetric key algorithm, AES 256, has been used. Symmetric keys can be pre-
selected by the user. The socket encryption functionality for received and transmitted data are also done
in Python, as is the rest of the framework. The Crypto library of Python is used to generate random
values employing the generator implemented in the Random module and all the related functionality for
encryption and decryption in AES.
The encryption function receives the parameter data, which contains a received data stream. At the
same time, decrypt receives the data encrypted for decryption and manipulation by the various process-
ing and malware detection modules of SecBrain.
Padding is used as a filler system in each encrypted message. The aim is to hide identical patterns
between different messages, eventually leading to security problems when carrying out statistical stud-
ies on the information exchanged.
The framework implementation also guarantees the user’s privacy, protecting data visualization from
entities external to the BCI headset and framework communication. The proposal has been identified as
a feasible and necessary solution that can be used in addition to existing protocols such as LSL.
Once the signal has been captured, the framework pre-processes it. This phase is crucial because once the
raw data has been acquired, it is practically unreadable. Thus, it is necessary to use a series of methods to
help make it easier to detect the P300. First, a noise filter is applied to the frequency 50Hz. Interference
189
SecBrain
caused by the power supply on this frequency is avoided (the power supply operates at 50Hz in Europe).
The next step is to apply a bandpass filter to limit the frequencies at which the P300 potential appears.
A bandpass filter is used with the IIR method in the frequency range of 1-17 Hz. The documentation
provided by the literature for its application has been followed, and the objective of obtaining a clean
signal without eliminating vital information to capture the P300 potential.
To supplement the signal processing, ICA is used to separate the components that make up the EEG
signal individually. The primary purpose is to reduce further the effect of noise on the data obtained. ICA
suppresses those artifacts that remain in the signal after applying the previous filters (notch, bandpass)
and affect the different channels, such as the flickers made during the experiment. Its operation is based
on finding projections with high non-gaussianity. In this way, a decomposition of the signal is obtained,
obtaining independent components.
Finally, to make the signal easier, a reduction of the signal samples is made (also called downsam-
pling). The framework reduces the number of samples for each second. The downsampling ratio used to
the EEG signal in this experiment has been 5. The signal will go from having 250 samples every second
(250 Hz) to 50 samples every second (50 Hz).
Figure 7 shows a fragment of the same recently captured EEG signal and the one resulting from ap-
plying all the processing functions mentioned in this subsection.
Figure 7. EEG signal before and after applying all the functions of the signal processing component
Although this process is a critical step in the BCI cycle functioning, the focus will be on the security
risks involved. As in the previous phase, data can be compromised by third-party software or malware.
For this reason, it is necessary to apply methods that offer integrity and confidentiality and prevent data
from being stolen or modified.
It is necessary to check that the hash corresponds to the one generated in the acquisition phase. If it
does not match, processing will be completed, and an alert will be displayed for a possible attack.
190
SecBrain
In general, most applications involving BCI devices are aimed at detecting the P300, creating robust
processes with various utilities. With its detection, it allows the development of robust applications for
a variety of purposes. For this reason, it is essential to determine if one or more P300 potentials have
been detected but even more so to determine if the signal has been compromised by introducing or at-
tenuating these potentials. Therefore, the fundamental solution principle is to make a good and reliable
detection of the P300 in specific EEG signal segments. In this way, SecBrain detects any deliberate
manipulation of the signal that could affect the application functionality and, consequently, the device
or person using the BCI.
SecBrain makes use of classifiers to carry out an automatic process and address this functionality.
Classifiers belong to supervised computer-based learning that attempts to predict the outcome based
on previous training models. If extrapolating the above definition to detect P300, these are classifiers
explicitly trained to detect P300.
SecBrain uses several classifiers with specific correlation analysis algorithms for the detection of
evoked potentials: Classifier I, use of scalar standardization algorithms and regressions; Classifier II, a
model with a linear decision limit, generated by adjusting conditional class densities to the data and us-
ing Bayes’ rule; Classifier III, the same operation as Classifier II but adding xDAWN as a spatial filter;
Classifier IV, estimation of the covariance matrix of the possible potentials, spatial projection of the
tangent and regressions; and Classifier V, with an analysis of the covariance matrix and classification
by Minimum to Medium Distance.
A manual division of the labeled EEG signal data is made into two different sets to train the classifiers:
training data and test, with proportions of 75% and 25%, respectively. The training data set (75%) has
been cross-validated and stratified (due to the unbalance dataset). The StratifiedShuffleSplit generator
(StratifiedShuffleSplit, 2020) determines the cross-validation split strategy. This strategy allows making
ten partitions of the input data, forming ten different combinations. Each combination is divided into two
data sets: training data and test data (with the same proportions as the previous division). While the first
ones are used to train the classifiers, the latter is used to evaluate the accuracy of the given predictions.
Figure 8 shows the supervised learning scheme followed. At this point, the cross-validator returns a set
of scores obtained for each iteration and combination generated (10 scores per qualifier).
191
SecBrain
Having trained models allows identifying P300 evoked potentials in any experiment carried out au-
tomatically. Together with appropriate security measures, such as those implemented in SecBrain, the
correct use of the BCI and the functionality it offers can be guaranteed.
Experiments
This section details a real experiment carried out to test the functionality of the SecBrain framework.
After that, it will analyze that each phase of the framework has carried out its task correctly. In other
words, it will check that a correct acquisition, processing, and detection of the P300 is performed.
The performed experiment follows the Oddball Paradigm technique to o2btain the P300 (Section
Use case). The images used in the video creation are random and without any special meaning for the
subject. They belong to the categories of landscapes, animals, or unknown subjects. One of them is ob-
tained as a target image and is shown at the beginning of the video. Then, all the photos are organized,
guaranteeing a brief and similar interval between them. The Target image is the only one that is repeated
many times randomly during the display. Table 3 includes the parameters used in the construction and
deployment of the experiment.
Three different experiments are carried out, in all of which the parameters mentioned above are applied.
In each experiment, wholly random and additional images are added. Two of them were performed on
the same subject, while the other was performed differently. The first individual was 22 years old, while
the second was 23 years old. Both were similar in height, approximately 1.80m. An external monitor
(Figure 1) at the subject’s eye level was used for visualization. The participants posture was upright and
perpendicular to the floor, avoiding involuntary movements or an uncomfortable position. Any distrac-
tion affecting the conditions during the recording is avoided. Both individuals were asked to focus on
the screen, if possible, at a fixed point on the monitor. Finally, once the images displayed were complete,
subjects were asked to say aloud the total number of images identical to that initially displayed and that
they had counted mentally during the session.
It should be noted that during the experiment, SecBrain generates logs with the internal structure that
is created to store the images in memory, the status of the security policies used during transmission as
192
SecBrain
well as any error or alert during the acquisition process. From the beginning of the experiment, the data
will be exchanged between the different components of the framework, producing the detection of the
P300 always within the generated security perspective. Therefore, the results will be reliable, consider-
ing that no type of attacker has affected the data.
Once the experiment details have been described, it will be checked whether the framework has
worked correctly. For this purpose, the Area Under the Curve (AUC) metric is used. This metric returns
a score for each classifier indicating how well or poorly the classifier predicts new P300s that the clas-
sifiers have not used for training. Finally, the AUC values are averaged by the classifier, obtaining an
average value AUC representing the classifier performance. Figure 9 shows the AUC value obtained
when supplying the test assembly to each classifier.
The training results have been outstanding (Figure 8); it can detect P300 with an acceptable percentage
(around 80% of AUC) of the P300 generated thanks to the set of images shown during the acquisition
phase. Of all the classifiers, classifiers I and V (with AUC values of 0.812 and 0.817, respectively) stand
out as the most promising.
The BCI trend is to have an interconnected world where many BCI will share thoughts, feelings, or
experiences. That is why cybersecurity is essential in the field of BCI. Possible future research could
raise new malware applications with different behaviors whose purpose is to determine the accuracy in
the detection of attacks. Another possibility would be to improve the existing infrastructure to provide
scenarios with more sophisticated perimeter security scenarios.
The proposed solution provides a significant advance in cybersecurity on BCIs, which is currently
almost non-existent. However, the viability of the future is notoriously low. The growth of distributed
attack vectors and robust implementations in terms of offensive attacks means that potential attackers
could overtake the proposed framework security.
193
SecBrain
Therefore, future research could pose new malware applications with different dynamics whose purpose
is to determine the accuracy of threat detection. The use of more current malware-based attacks could
be used to determine the framework potential for detection. Similarly, it could involve a renewal of the
system to adapt to new requirements. Another possibility would be to improve the existing infrastructure
to provide more sophisticated perimeter security scenarios. Therefore, cybersecurity opens an opportunity
to enhance the BCI hardware to offer new types of more secure connections or the possibility of storing
or generating cryptographic keys.
CONCLUSION
The benefits provided by BCIs are very varied, ranging from improving people’s well-being to enter-
tainment and leisure and security by providing new authentication techniques. However, these devices
continuously deal with such compromised user’s data and certain security aspects. The most relevant
elements are mainly centered on the theft of sensitive data or the BCI control and its applications. In
this context, the main contribution of this chapter is to design and implement an intelligent and secure
framework, called SecBrain, for the detection of P300s generated by visual stimuli. For this purpose,
different techniques are considered by SecBrain, being the first one to apply a hash mechanism to both
the data used for image sampling and the data received from the EEG.
Additionally, an attacker could modify the EEG signal before reaching the machine in charge of pro-
cessing the signal and detect the P300. For this purpose, SecBrain implements encryption mechanisms
to the sockets that exchange the data, thus preventing a possible leakage of information. The SecBrain
framework will automate these techniques to detect any threat that could compromise the BCI life cycle,
deteriorating the user experience. With all this, security mechanisms are proposed to guarantee the safe
use of the BCI devices and allow a first approach to a more reliable and robust infrastructure.
ACKNOWLEDGMENT
This work has been partially supported by (a) Bit & Brain Technologies S.L. under the project Cyber-
Brain, associated with the University of Murcia (Spain), by (b) the Swiss Federal Office for Defense
Procurement (armasuisse) with the CyberSpec (CYD-C-2020003) project, and by (c) the University of
Zürich UZH.
REFERENCES
Ahn, M., Lee, M., Choi, J., & Jun, S. (2014). A Review of Brain-Computer Interface Games and an
Opinion Survey from Researchers, Developers, and Users. Sensors (Basel), 14(8), 14601–14633.
doi:10.3390140814601 PMID:25116904
Ajrawi, S., Rao, R., & Sarkar, M. (2021). Cybersecurity in Brain-Computer Interfaces: RFID-based design-
theoretical framework. Informatics in Medicine Unlocked, 22, 100489. doi:10.1016/j.imu.2020.100489
194
SecBrain
Al-Nuaimi, F. A., Al-Nuaimi, R. J., Al-Dhaheri, S. S., Ouhbi, S., & Belkacem, A. N. (2020). Mind Drone
Chasing Using EEG-based Brain Computer Interface. 2020 16th International Conference on Intelligent
Environments (IE), 1. 10.1109/IE49459.2020.9154926
Arrichiello, F., Di Lillo, P., Di Vito, D., Antonelli, G., & Chiaverini, S. (2017). Assistive robot oper-
ated via P300-based brain computer interface. 2017 IEEE International Conference on Robotics and
Automation (ICRA), 1. 10.1109/ICRA.2017.7989714
Aydemir, O. (2015). Improving classification accuracy of EEG based brain computer interface sig-
nals. 2015 23nd Signal Processing and Communications Applications Conference (SIU), 1. 10.1109/
SIU.2015.7130442
Badcock, N. A., Mousikou, P., Mahajan, Y., de Lissa, P., Thie, J., & McArthur, G. (2013). Validation of
the Emotiv EPOC® EEG gaming system for measuring research quality auditory ERPs. PeerJ, 1, e38.
doi:10.7717/peerj.38 PMID:23638374
BCI Lab | Universidad de Granada. (2020). BCI Lab. http://www.ugr.es/%7Ebcilab/
Chaudhary, P., & Agrawal, R. (2018). Emerging Threats to Security and Privacy in Brain Computer
Interface International Journal of Advanced Studies of Scientific Research, 3(12). https://ssrn.com/
abstract=3326692
Chiappa, K. H., & Ropper, A. H. (1982). Evoked Potentials in Clinical Medicine. The New England
Journal of Medicine, 306(19), 1140–1150. doi:10.1056/NEJM198205133061904 PMID:7040957
Choo, A., & May, A. (2014). Virtual mindfulness meditation: Virtual reality and electroencephalography
for health gamification. 2014 IEEE Games Media Entertainment, 1. doi:10.1109/GEM.2014.7048076
Crea, S., Nann, M., Trigili, E., Cordella, F., Baldoni, A., Badesa, F. J., Catalán, J. M., Zollo, L., Vitiello,
N., Aracil, N. G., & Soekadar, S. R. (2018). Feasibility and safety of shared EEG/EOG and vision-guided
autonomous whole-arm exoskeleton control to perform activities of daily living. Scientific Reports, 8(1),
1. doi:10.103841598-018-29091-5 PMID:30018334
Curran, M. T., Yang, J., Merrill, N., & Chuang, J. (2016). Passthoughts authentication with low cost
EarEEG. 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology
Society (EMBC), 1979-1982. 10.1109/EMBC.2016.7591112
Donchin, E. (1979). Event-related Brain Potentials: A Tool in the Study of Human Information Process-
ing. Evoked Brain Potentials and Behavior, 13–88. doi:10.1007/978-1-4684-3462-0_2
EEGLAB. (2020). https://sccn.ucsd.edu/eeglab/index.php
Finke, A., Lenhardt, A., & Ritter, H. (2009). The MindGame: A P300-based brain–computer interface
game. Neural Networks, 22(9), 1329–1333. doi:10.1016/j.neunet.2009.07.003 PMID:19635654
Fu, H.-L., Fang, P.-H., Chi, C.-Y., Kuo, C., Liu, M.-H., Hsu, H. M., Hsieh, C.-H., Liang, S.-F., Hsieh,
S., & Yang, C.-T. (2020). Application of Brain-Computer Interface and Virtual Reality in Advancing
Cultural Experience. 2020 IEEE International Conference on Visual Communications and Image Pro-
cessing (VCIP), 351–354. 10.1109/VCIP49819.2020.9301801
195
SecBrain
Ghoneim, A., Muhammad, G., Amin, S. U., & Gupta, B. (2018). Medical image forgery detection for
smart healthcare. IEEE Communications Magazine, 56(4), 33–37. doi:10.1109/MCOM.2018.1700817
Gupta, B. B., Perez, G. M., Agrawal, D. P., & Gupta, D. (2020). Handbook of computer networks and
cyber security. Springer. doi:10.1007/978-3-030-22277-2
Gupta, C. N., Palaniappan, R., & Paramesran, R. (2012). Exploiting the P300 paradigm for cognitive
biometrics. International Journal of Cognitive Biometrics, 1(1), 26. doi:10.1504/IJCB.2012.046513
Iturrate, I., Antelis, J. M., Kubler, A., & Minguez, J. (2009). A Non-invasive Brain-Actuated Wheelchair
Based on a P300 Neurophysiological Protocol and Automated Navigation. IEEE Transactions on Robot-
ics, 25(3), 614–627. doi:10.1109/TRO.2009.2020347
Jang, Y. S., Ryu, S. A., & Park, K. C. (2011). Analysis of P300 Related Target Choice in Oddball Para-
digm. Journal of Information and Communication Convergence Engineering, 9(2), 125–128. doi:10.6109/
jicce.2011.9.2.125
Kaongoen, N., Yu, M., & Jo, S. (2020). Two-Factor Authentication System Using P300 Response to a
Sequence of Human Photographs. IEEE Transactions on Systems, Man, and Cybernetics. Systems, 50(3),
1178–1185. doi:10.1109/TSMC.2017.2756673
Letswave7 | Letswave.cn. (2018, July 17). https://letswave.cn/
Liberati, G., da Rocha, J. L. D., van der Heiden, L., Raffone, A., Birbaumer, N., Olivetti Belardinelli,
M., & Sitaram, R. (2012). Toward a Brain-Computer Interface for Alzheimer’s Disease Patients by Com-
bining Classical Conditioning and Brain State Classification. Journal of Alzheimer’s Disease, 31(s3),
S211–S220. doi:10.3233/JAD-2012-112129 PMID:22451316
López Bernal, S., Huertas Celdrán, A., Martínez Pérez, G., Barros, M. T., & Balasubramaniam, S. (2021).
Security in Brain-Computer Interfaces: State-of-the-Art, Opportunities, and Future Challenges. ACM
Computing Surveys, 54(1), 35. doi:10.1145/3427376
Martinovic, I., Davies, D., & Frank, M. (2012). On the feasibility of side-channel attacks with brain-
computer interfaces. Proceedings of the 21st USENIX Security Symposium, 143-158.
MATLAB - El lenguaje del cálculo técnico. (2020). MATLAB & Simulink. https://es.mathworks.com/
products/matlab.html
Moreno, I., Batista, E., Serracin, S., Moreno, R., Gómez, L., Serracin, J., Quintero, J., & Boya, C.
(2019). Los sistemas de interfaz cerebro-computadora basado en EEG: características y aplicaciones.
I+D Tecnológico, 15(2), 13–26. doi:10.33412/idt.v15.2.2230
Na, E., Lee, K., Kim, E. J., Bae, J. B., Suh, S. W., Byun, S., Han, J. W., & Kim, K. W. (2021). Pre-
attentive Visual Processing in Alzheimer’s Disease: An Event-related Potential Study. Current Alzheimer
Research, 17(13), 1195–1207. doi:10.2174/1567205018666210216084534 PMID:33593259
Nurseitov, D., Serekov, A., Shintemirov, A., & Abibullaev, B. (2017). Design and evaluation of a P300-
ERP based BCI system for real-time control of a mobile robot. 2017 5th International Winter Conference
on Brain-Computer Interface (BCI), 1. 10.1109/IWW-BCI.2017.7858177
196
SecBrain
OpenViBE | Software for Brain Computer Interfaces and Real Time Neurosciences. (2020, December
10). OpenViBE. http://openvibe.inria.fr/
Polich, J., & Heine, M. (1996). P300 topography and modality effects from a single-stimulus paradigm.
Psychophysiology, 33(6), 747–752. doi:10.1111/j.1469-8986.1996.tb02371.x PMID:8961797
Polich, J., Howard, L., & Starr, A. (1985). Effects of Age on the P300 Component of the Event-related
Potential From Auditory Stimuli: Peak Definition, Variation, and Measurement. Journal of Gerontology,
40(6), 721–726. doi:10.1093/geronj/40.6.721 PMID:4056328
Rathi, N., Singla, R. & Tiwari, S. (2021). A novel approach for designing authentication system using
a picture based P300 speller. Cogn Neurodyn. doi:10.1007/s11571-021-09664-3
Simon, A. J., Bernstein, A., Hess, T., Ashrafiuon, H., Devilbiss, D., & Verma, A. (2011). P1-112: A brain
computer interface to detect Alzheimer’s disease. Alzheimer’s & Dementia, 7(4S_Part_4), S145–S146.
doi:10.1016/j.jalz.2011.05.391
StratifiedShuffleSplit — scikit-learn 0.24.0 documentation. (2020). Scikit-Learn. https://scikit-learn.
org/stable/modules/generated/sklearn.model_selection.StratifiedShuffleSplit.html
Sundararajan, K. (2017). Privacy and security issues in Brain Computer Interface [Unpublished master’s
thesis]. Auckland University of Technology, Auckland, New Zealand.
The OpenBCI GUI · OpenBCI Documentation. (2020). OpenBCI. https://docs.openbci.com/
docs/06Software/01-OpenBCISoftware/GUIDocs
Unsplash. (2020). Unsplash API Documentation. https://unsplash.com/documentation
Wronka, E., Kaiser, J., & Coenen, A. M. (2008). The auditory P3 from passive and active three-stimulus
oddball paradigm. Acta Neurobiologiae Experimentalis, 68(3), 362–372. PMID:18668159
197
SecBrain
198
199
Chapter 11
A Study on Data Sharing Using
Blockchain System and Its
Challenges and Applications
Santosh Kumar Smmarwar
National Institute of Technology, Raipur, India
Govind P. Gupta
https://orcid.org/0000-0002-0456-1572
National Institute of Technology, Raipur, India
Sanjay Kumar
National Institute of Technology, Raipur, India
ABSTRACT
Blockchain since 2009 has been gaining more popularity in various fields to use in numerous applications
to overcome the security issues such as privacy, transparency, and mutability of data in the process of
data sharing. Process of data sharing has many addressed and unaddressed challenges such as infor-
mation encryption and decryption, data authentication, storage security, latency time, transfer speed of
data, detecting malicious nodes, prevent the computer system from attacks, trust in the sharing process.
In this chapter, the authors have reviewed the data sharing paper based on blockchain technology and
presented the analysis of various techniques used in the information sharing process. The comprehensive
analysis is categorizing in the following areas like incentive mechanism-based work, IoT-based data
sharing, healthcare data sharing, and internet of vehicle data sharing using blockchain.
INTRODUCTION
In the recent era of Big Data development with the advancement in information and communication
technology the rate of growing data and digital resources has been grown exponentially. Storing these
data has become important for customer and business organizations (Feng et al., 2019). One of the ways
to control the growth rate of data to reduce the same information creation is through the data sharing
DOI: 10.4018/978-1-7998-7789-9.ch011
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
process in various organizations such as health care, academic, financial sector, etc. Data sharing has many
uses in various fields such as healthcare, the internet of things, smart devices, and internet of vehicles,
supply chains, and logistic networks. In data sharing, secure data sharing is necessary to achieve security
goals like confidentiality, integrity, availability, authentication, transparency, and privacy of the data.
These problems can be overcome by using blockchain technology in making data sharing transparent,
immutable, decentralized storage, integrity protected, and confidentiality. The feature of transparency and
immutability of blockchain made it most popular to deploy or integrate into secure data sharing. Block-
chain technology is based on the three pillars that are distributed ledger, peer-to-peer communication,
and consensus protocol. These are the core of this technology that makes it more secure and transparent
(McGhin et al., 2019). In this survey study, we have reviewed the blockchain used data sharing papers
and analyses various methods used by the early researcher for secure data sharing to resolve the privacy
concern by using some cryptographic techniques.
OUR CONTRIBUTION
In this research paper we have reviewed the work of data sharing using blockchain technology, to get a
research trends of various security mechanism used in blockchain for information sharing. We found that
in our study most of the author described comprehensive overview of secure data sharing by proposed
method such as incentive mechanism, privacy preserving scheme, dynamic data access control policy,
secret sharing schemes and some attribute based encryption technique. In this paper we demonstrated
the summary of existing work and methods used in blockchain to secure data sharing. Shown existing
work, pros and cons of early research work in areas of healthcare data sharing and IoT device-based data
sharing, incentive-based data sharing and various industrial application.
MOTIVATION
As mentioned above security challenges faced by centralized authority such as single point of failure may
lead to the collapse of all data, high sharing cost, reduce manual verification (Goyal et al., 2018), reduce
wastage of time, brings transparency and immutability in the system encourage us to work on secure data
sharing among the peer to peer or decentralize environment by integrating blockchain technology and
smart contact functionality (Mohanta et al., 2019). Since the inception of blockchain technology came
into existence these have brought some potential solutions that may be overcome by using blockchain
in various emerging fields of information technology.
Blockchain is distributed ledger technology initially developed for cryptocurrency purposes like bitcoin
by an unknown person named Satoshi Nakamoto in 2009(Feng et al., 2019), (McGhin et al., 2019). It is
a decentralized system of digital ledger stored the synchronized copy of data. The blockchain system is
based on asymmetric cryptography where two keys, a public key for device identification and a private
key for signing transactions used (Goyal et al., 2018),. Blockchain is a chain of the block, each block
200
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
linked to the next block via hash value in chronological order. Blockchain is supported by a combination
of three technologies that are distributed ledger technology, peer-to-peer network, and consensus proto-
col among all nodes (Mohanta et al., 2019). In decentralized, ledger block is added through consensus
mechanism among nodes in the blockchain network. There is some of the popular consensus protocol
used in blockchain mining as proof of work (POW), Proof of stake (PoS), and Practical Byzantine Fault
Tolerance (PBFT) protocol. This is the most popular technology in recent times used by industry and
academia nowadays. It has the characteristics of transparency, immutability, decentralization, anonym-
ity (Hasselgren et al., 2019). The simple blockchain diagram has depicted below where every block is
added from its previous block through hash value in a chronological way.
CLASSIFICATION OF BLOCKCHAIN
• Public Blockchain: Public blockchain is decentralizing networks to have more security. In this,
ledgers are visible to all users who are connected to public networks. Anyone is free to join the
network and add a block to the ledger (Xu et al., 2017). The most frequently used example is bit-
coin and ethereum.
• Private Blockchain: In a private blockchain only selected nodes are allowed to join the network
and add the data to the digital ledger. This allows anyone to view the ledger of data for example
hyperledger fabric (Dinh et al., 2017).
• Consortium Blockchain: It is a combination of a group of the organization located at a different
location (Dib et al., 2018). Only predefined organizations communicate, verify, and append-only
transactions to the distributed digital ledger (Gai et al., 2019).
PROPERTIES OF BLOCKCHAIN
There are the following properties of blockchain technology that make it a powerful system to use in
various fields as a solution (Liu et al, 2019).
201
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
• Autonomous: Autonomous property of blockchain makes it free from third party’s control.
Anyone can enter into the network verify and append transactions to the blockchain database.
• Distributed: in this, once the block confirms by all the peer’s entities by some proof of mecha-
nism then it is added only to the ledger. It works based on peer-to-peer infrastructure.
• Immutability: Immutability means no modification is possible in the block, once the block hap-
pened to the ledger. It would reflect all nodes in the global ledger.
• Decentralize: decentralization property makes the elimination of third parties involved in pro-
cessing the transaction, reduces the cost, time and brings transparency among the nodes.
• Anonymity: this property hides the user identity in the blockchain network so that no node in the
network trusts other nodes or its peer node.
Blockchain technology has potential industrial application in many fields such healthcare industry, smart
transportation, finance sectors, E-governance, food supply chain and many more in data sharing to brings
the transparency, immutability, traceability and integrity as well. Blockhain technology can be useful in
reducing the cost of production by using integrated smart contract facility and eliminates the central control
of data. In healthcare sector it may provide the secure access of data sharing with greater transparency
among different hospital globally to provide quick treatment to the patients. The finance sector can be
achieved with less number fraudulent activities by using blockchain in transactional process. In smart
transportation blockchain can support transparent, secure tracking of vehicles and brings transparency
in insurance work to prevent illegal claim. The E-governance sector suffered from lack of transparency,
so it also has wide applicability in this sector to provide secure authentication, data privacy and prevent
fake data auditing. Similarly for other sectors also blockchain has various industrial application in terms
of transparence, security, decentralization storage, integrity and immutability (Dubovitskaya et al., 2020).
202
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
LITERATURE SURVEY
In this survey of data sharing based on a blockchain platform, we have comprehensively reviewed papers
on data sharing based on cloud data computing, Internet of Things (IoT), healthcare data, internet of
vehicles, and blockchain-based platform for secure data sharing.
In paper (Bhaskaran et al., 2018) the work of the author has based on a user-associated consent-driven
mechanism, in which a double-blind data-sharing approach is used for KYC (know your customer)
validation using a blockchain system for banking institutions. This provides the dynamic access control
of data sharing. It reduces the time and cost of manual verification. It has some drawbacks such as scal-
ability issues, limited storage, and low transactional throughput.
203
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
In the paper (Shrestha et al., 2018) author proposed a decentralized data sharing framework based
on blockchain for incentivizing data owners with digital tokens, author introduced blockchain with a
smart contract concept for incentivizing participants in sharing their data and reward digital tokens.
This approach of data sharing allows only the interested data seekers and ensures the data owner of
misusing their data by illegal users. In this model, the author brought the escrow service to manage the
user with legal obligations. The proposed framework is designed with due care of transparency, access
control policy and specifies the purpose of data sharing. In paper (Naz et al., 2019) author focused on
data authenticity and data quality during the process of information sharing. To achieve this author
used the Shamir’s secret sharing encryption technique and the online review system of Watson analyzer
to identify the fake reviews. The advantage of this work is no leakage of hashed information, lowest
computation time in encryption.
In paper (Wang et al., 2019) also author proposed an incentive scheme that discussed boosting
blockchain-based IoT data sharing of the historical block. In this work, the author used the coin-locking
strategy and micro-payment system to speed up the transmitting process. To ensure the security of data
Shamir’s secret sharing scheme played a great role. This proposed work can identify the malicious or
dishonest nature of the sharing nodes. Although this system is efficient it has some high bandwidth cost
limitations.
In the paper (Xuan et al., 2020) author presented an incentive mechanism for secure data sharing.
This incentive method is based on evolutionary game theory which encourages participants to share
their data securely and establish trust among the user who wants to share data to the blockchain-based
secure platform. In addition to the game, theory the author applied the evolutionarily stable strategy
and symmetric game theory to predict the most trustable sharing node among the various participants.
However, this method has some limitations of data size and quality of data obtains by the recipient.
In (Wang et al., 2017) author addressed the problem of conventional resource sharing in government
departments such as efficiency, reliability, and security. These issues could be solved by the emerging
blockchain technology. In this work, the author proposed the blockchain concept on government informa-
tion, asset detail, resource sharing system to improve its reliability and security through the features of
transparency, immutability, and traceability, and decentralization network of blockchain. This technol-
ogy reduces the cost of implementation of sharing process than the traditional approach. However, this
technology needs to be enhanced for the large scalable network in government setup and optimize the
efficiency as well as the performance of the system. The work (Gupta & Gupta, 2017) has been discussed
about the web application threats and vulnerabilities that may cause the major disruption in using the
web application among various organizations. The proposed work concerns the potential security flaws
and measures the threat.
In paper (Nakasumi et al., 2017) author presented the solution to solve the problem of double mar-
ginalization and information inconsistency in supply chain management. This system uses blockchain
and homomorphic encryption techniques to provide transparency, immutability, traceability to user data.
In this system author also added legal and authentic regulatory decisions of collecting data, storing, and
sharing sensitive information. With the use of blockchain technology at supply chain it minimizes complex
problem such as capacity risk in supply chain and improves the supply chain efficiency. It also increases
204
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
the demand forecasting of consumers up to some extent. In the future, this solution needs to be bring-
ing some incentive mechanism to handle the capacity risk and encourage users in sharing information.
In paper (Zhang et al., 2018) author focused on the privacy threat of data stored on the cloud due
to the lack of processing power and storage capacity of IoT devices. To overcome these threats of data
stealing and security author proposed architecture combined with the functionality of privacy preser-
vation and authorized fine-grained user control access. In this work author used two types of method
that is the attribute-based signature (ABS) and ciphertext-policy attribute-based encryption (CP-ABE).
The ABS is used to provide fine-grained access control that identifies users or devices and attribute of
signature instead of other signature schemes like ECDSA (Elliptic curve digital signature algorithm).
In the paper (Wang et al., 2018), the author discussed issues in existing data sharing models such as
transparency and traceability methods. To improve these issues author presented the blockchain tech-
nology as a new data-sharing scheme. In which the author introduced the concept of the double-chain
structure of blockchain, these double structure concepts differentiate between original data storage and
transaction data. The first structure is responsible for storing original data and the second is for transac-
tion data. To make reliable and safe sharing author used the proxy re-encryption technique.
In (Rawat et al., 2018) author presented the information sharing framework for cybersecurity infra-
structure to prevent, detect, and respond to cyber attackers. The Cyberspace world is facing the ongoing
challenge of safeguard confidential information and critical assets from potential future attacks. The
proposed architecture known as ishare uses blockchain technology to share cyber threat information
securely, reliably, and transparently among multiple organizations. The ishare framework is captured
constantly high-resolution cyber-attack information to prevent IT resources from being compromised by
an attacker. The proposed framework aimed at providing the procedure of transaction in ishare, cyber-
attack detection and sharing information, deploying cyber-defense solutions and updates. To analyzes,
the proposed framework author used the one-way-attack, two-way attack, and Stackelberg game for
cyber-attack and defense analysis approach for security performance.
In (Cash et al., 2018) author presented a two-tier blockchain model consist of permission and
permission-less blockchain to analyze the computing power of proof of work consensus algorithm and
Proof of Authority protocol algorithm on ethereum platform. The environment used in work is Mac OS
and Ubuntu Linux operating system. the result analysis shows the better performance of PoA algorithm
on permissioned blockchain in term of constant or no change in block count on both operating systems
when the number of nodes increases. In permission, blockchain required less computation power for
consensus algorithm than permission-less. However, the less computation of permissioned blockchain
may possess the threat of malicious users due to its less computing power required by consensus protocol.
So this needs to establish reliable and strong trust among nodes.
In paper (Wang et al., 2019) author addressed the problem of data sharing flexibility where some of
the public-key encryption with access control (PEAC) schemes are not able to control the encryption
easily on both sides by the data owner. So in this work author proposed the functional broadcast encryp-
tion (FBE) to control the file-based encryption and receiver-based encryption simultaneously. This FBE
technique can share the set of the file to a group of users. FBE is an expansion of the PEAC scheme. It
is more efficient, flexible, feasible in terms of technically, economically and cloud storage performance
while PEAC has storage overhead costs.
In (Shrestha et al., 2019) author has shown the concern over online services used by a user in the
travel domain, addressed the privacy concern of user’s data collected by travel agencies. In this paper,
the author proposed the framework of user-controlled privacy-preserving to user’s data while using
205
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
online services of hotel booking systems through blockchain technology. This model is developed on a
multichain framework to control the user profile data like name, nationality, birth date, contact phone
number, address of the user, and purpose of travel. The multichain framework relies on the public key
encryption technique. Multichain is used to restrict access to data by authorized users only. Multichain
is a kind of private blockchain to maintain the privacy and access policy of data
In (Wu et al., 2019) author presented the efficient, effective, and user-controlled privacy-preserving
traceable attribute-based encryption scheme in blockchain environment to protect the data integrity and
non-repudiation of transactions. The data-sharing environment poses some security threats of cracking
the secret key and leakage of sensitive information in a distributed environment, so to prevent these
challenges author proposed the pre-encryption technology to improve the efficiency of attribute-based
encryption (ABE), this pre-encryption technique does the necessary pre-calculation before the message
to be encrypted is well known. The attribute bloom filter (ABF) is used in this scheme to determine the
existence of an attribute or element in the set and hide the attributes in an anonymous access control
scheme. To prevent the cracking of the secret key author combined the user’s signature and the main
master secret key of Attribute Authority in the user’s master secret key.
In (Samuel et al., 2019) author presented the privacy and security concern to achieve efficient services
by service providers in data sharing. Because huge data is generated by multiple smart home appliances
that possess privacy and security threats and lacking fair data sharing which reduces the transparent
participation of users. so given the author of the above issue proposed a fair data sharing scheme to
encourage user participation based on their reputation score by using the PageRank mechanism in this
work, the Pagerank mechanism provides the authenticity of node and adds block into the ledger.
In paper (Eltayieb et al., 2020) author has pointed out the drawback of the traditional cloud storage
system of data sharing that poses the service availability, centralize information database, high running
operational cost, and privacy concerns. To come out of this concern author proposed the integrated
concept of blockchain consists of attribute-based encryption to make the secure cloud environment for
information exchange. In this survey work, the concept of a smart contract is used to provide storage
efficiency in the cloud atmosphere. In this work, the author achieved the confidentiality and integrity
of data by using the concept of a secret sharing scheme and access tree.
In (Rahman et al., 2020) author proposed a framework based on Accountable cross-border data shar-
ing, in this work author, presented a global cloud platform connected to the security gateway of different
regions that allows data sharing among different countries securely. This framework has the feature of
penalty for misbehaving nodes or entities in the sharing process. The authenticity of data is verified by the
Elliptic curve digital signature algorithm. The framework consisted of three entities such as data sender,
data receiver, and any entity or party. This platform allows the sharing based on the under relaxed trust
assumption on sender and receiver. The author in (Sumathi, & Sangeetha, 2020) provides the concepts
of blockchain in banking system to make efficient and secure storage with distributed ledger property
of blockchain. Author highlights the risk of data center failure can be overcome by using decentralize
system of blockchain and it provides the immutability in data modification. The author in (Mohan, &
Gladston, 2020) proposed the work for cloud data auditing with help of merkle tree and blockchain to
maintain the integrity, transparency and immutability. The blockchain store transaction in distributed
form, so that it is difficult to modify the data by unauthorized users.
206
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
In the paper (Amofa et al., 2018) author has shown challenges for health information exchanges and
the inability to control the data once has been transmitted such as privacy and integrity. The problem of
data access control discourages the participants from sharing the health data. For this author proposed
a blockchain-based system framework for secure control and sharing data among different hospitals.
This is done by pairing-based user-generated acceptable use policies by using the smart contract. This
framework minimizes the threat to data after sharing. In this work, the author used the cryptographic
key method and smart contract to define the access level of a person’s data. By this approach, the author
reduced the financial cost of managing data, improve the efficiency of accessing health data, provide
secure distribution of data, and has low latency. However, this proposed model has scalability problems.
In (Guo et al., 2018) author address, the issues of authenticity of electronic health record (EHR)
placed inside on blockchain. So to prove the validity or authenticity of EHR, the author presented the
secured attribute-based ensured signature scheme with more than one authority. In this scheme, the se-
cret pseudorandom function seeds are used among the multiple entities that prevent the collision attack.
The computational bilinear Diffie-Hellman technique is used to achieve no modification and complete
privacy, the integrity of attribute signer. The ABS is secure in the random oracle model. This scheme
improves the performance and cost of medical data sharing among many authorities. The prime objec-
tive of the ABS scheme is to prove the validity of EHR data.
In (Wang et al., 2018) author presented the framework of a secure cloud-assisted EHR system by
using Attribute cryptosystem along with blockchain to reach confidentiality, integrity, authentication,
and fine-grained access control of data. This model integrates attribute-based encryption followed by
identity-based encryption (IBE) to encrypt the patient secret data. The identity-based signature (IBS)
is used to provide the digital signature for the authenticity of EHR data. This scheme combined all the
techniques into a single one to achieve different functionality this known as attribute-based/identity-
based encryption and signature(C-AB/IB-ES). Cloud storage provides fast transmission, file sharing,
data storing space, minimum cost, and efficient access.
In paper (Thwin et al., 2018) author analyzed the personal health record system (PHR) in regards to
blockchain properties that may arise the concern of privacy, integrity as well as confidentiality due to the
transparency property of blockchain. The author also described some of the other issues of blockchain
such as concern of storage, privacy aspect, user consent revocation, model performance, energy cost,
and scalability. To resolve some of the above-mentioned problems such as on-chain data privacy; limited
storage author used the Shamir’s secret sharing data scheme by using the concept of proxy re-encryption
technique to ensure the privacy concern of PHR. the authenticity is ensured by verified signature before
storing data on the blockchain and all the data are inclusive with the signature of gateway server on the
blockchain. This model also considered the availability of data by storing only meta-data on blockchain
nodes and privacy concerns by using a fine-grained access control policy. In this access control policy,
the encryption is done only by the data owner by using proxy re-encryption and the keys are known to
the gateway server.
In paper (Liu et al., 2018) author focused on the concern regarding privacy and security of files stored
on the cloud server. For this author used ciphertext-policy attribute-based encryption (CP-ABE). The
CP-ABE provides a secure access control policy to a user, a secure decentralized environment, and pro-
vide user-friendly service. The confidentiality of data and efficiency of the model are guaranteed by the
symmetrical encryption algorithm. The secure access control is provided by the CP-ABE technique and
207
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
with an access tree structure. It also ensures anti-collision attack and data integrity by using the crypto-
graphic hash value. This model has the advantage of the higher efficiency with symmetrical encryption
for files as well as CP-ABE fork, increased reliability, data security, and integrity. However, the model
has some limitation like computation time increase as the number of attribute increase in the policy.
In paper (Theodouli et al., 2018) author has pointed out the concern regarding privacy and security
needs in healthcare data sharing. For this concern author presented the potential use of blockchain to
address these issues through the smart contract feature of blockchain technology. The smart contract
further categorizes into three contracts for different functions such as registry contract for users regis-
try, patient data smart contract contains the hashed data of health information and permission contract
for access control of data. This work aimed at private data sharing and access permission of healthcare
information. The blockchain-based architecture has achieved the security goal while sharing healthcare
information such as integrity, user identity and provides accountability and auditing. However, this
architecture is not suitable for the large-scale network.
In (Liu et al., 2019) author aimed at the nature of medical data privacy and sensitivity during sharing
and provides a protection scheme for patient data. In this scheme, the author used private blockchain to
address the security and privacy concern. The private blockchain is more secure than the public block-
chain to ensure the security of electronic health records. This blockchain-based scheme satisfies security
requirements such as transparency, tamper proofness, immutability. In this system for mutual authen-
tication and generating session key author used the symptom matching algorithm for communication
of two same type disease patients about their illness in future. In this work, the Author used delegated
proof of stake (DPoS) mechanism to prevent dishonest nodes participation in the data sharing process.
In (Wang et al., 2019) author has shown the security threat to cloud-based stored medical data such
as privacy and security issues. In this regard author used blockchain technology as a solution to provide
data privacy and security to cloud-based stored information. The cloud manages for storing the EHR
ciphertext and hybrid blockchain contain the indexes of electronic health record (EHR). The framework
proposed is reliable, secure, and effective efficient privacy-preserving EHR sharing protocol by using
the searchable encryption and conditional proxy re-encryption technique. This framework uses the
proof of authority consensus protocol as the authenticity of the nodes in the decentralized network that
is consortium blockchain. The cryptographic primitives such as bilinear map, public-key encryption
with conjunctive keyword search technique, and conditional proxy re-encryption are implemented on
ethereum blockchain platform.
In (Nguyen et al., 2019) author discuss the concern of storage security of electronic health records(EHR)
on mobile-based cloud environment that having the security and privacy concern while sharing informa-
tion among patients, healthcare providers, and third party. Given the above issues, the author presented
a novel EHRs sharing model which integrates blockchain ledger and decentralizes interplanetary file
system (IPFS) with mobile cloud network. To provide secure access control and EHR sharing to medi-
cal data we use the smart contract. In this model proof of concept, the mechanism is used to provide
decentralized access control of EHR data sharing on mobile cloud framework to analyze the proposed
scheme. The advantage of this model is to provide the minimum network latency and data security as
compared to other existing information-sharing models and it is feasible for various e-health applications.
The work (Gupta et al., 2021) proposed the blockchain based secure concept for healthcare by using
the cloud assisted system and make the data sharing efficient by using the attribute based encryption
method. The proposed method is computationally efficient and storage efficient as well as robust.
208
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
In (Liu et al., 2018) author discussed the issues related to collecting high-quality data from IoT mobile
terminals and how securely share these data among mobile terminals, prevent device failure and com-
munication failure. To overcome these above challenges, blockchain technology is proposed for data
sharing and data collection efficiently. In this paper author combined ethereum platform with deep rein-
forcement learning to make an efficient, secure environment for data sharing and exchange. Reinforce-
ment learning facilitates the collection of a large amount of data while blockchain provides reliability,
security, and efficiently sharing data. Ethereum node stores the data and creates a private blockchain to
share data. This private network has two nodes named mining node and non-mining node. The mining
node checks the validity of blocks and adds them into the ledger while the non-mining node is used as
receiving and broadcasting e data sharing among nodes. The objective of this proposed system is to
support the mobile terminal to sense surrounding personal of interest devices to gain an edge in higher
data collection, geographic transparency, impartiality, and minimum power consumption. After the
simulation test, this scheme is feasible to higher security, reliability, and anti-collusion to DoS, DDoS
in data sharing. However, this system is not extendable to every blockchain node and executes multiple
tasks at the same time on each mobile terminal.
In paper (Si et al., 2018) author proposed the lightweight IoT information sharing security frame-
work by using decentralize platform. The author used the double chain model that combines the data
blockchain and transaction, which protects source data storage and is responsible for storing the indexes
of transaction data respectively. The double-chain concept is used to maintain data consistency and
avoid tampering with IoT data. The practical Byzantine fault-tolerant consensus mechanism increases
the registration efficiency of nodes, higher the transactional throughput and privacy of sharing data by
using the partial blind signature method. In This paper author used the dynamic game method which
provides the cooperation among the node and detect the malicious behavior of a node, this dynamic
game method find out the state of an unauthorized node and estimate the reputation value by using
PageRanker algorithm, higher reputation value provides trust in sharing the data. The loss of private
keys is secured by secret sharing.
In (Pham et al., 2019) author addressed the issues of time and cost in data collection from sensor
devices in IoT environments for intelligent systems. These data collections also have security challenges
like data leakage, privacy, and integrity. This author proposed the blockchain-based architecture which
ensures integrity, authorization, confidentiality, and transparency in the data sharing process. The data
collection is being done through the IoT gateway. After receiving data the IoT gateway encrypts these
data and stores them at off-chain storage, asymmetric encryption technique used on ethereum blockchain
ensuring data security and integrity to overcome these security issues. So that to get higher transaction
speed and provides reliable service availability. However, this has scalability concerns to include large
peers at the network.
In (Hofman et al., 2019) author addressed the demand of customers in manufacturer, retailer, and
supplier like food security and safety, sustainability. For this author proposed a methodological approach
in a complex organization such as supply chain and logistic network for development and deployment
of data sharing by using the blockchain technology with smart contract. This approach of data sharing
in supply chain and logistic networks use blockchain technology to reduce the implementation time in
commodity trading in the supply chain. In this work, the technology used by the author is distributed
209
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
ledger technology that creates data sharing reference model for supply and logistic network, informatics
principles as a Turing machine, ontologies
In the paper (Cech et al., 2019) author discussed the creation of a huge amount of data from IoT
devices and sensor nodes, collecting and securely sharing these data is the biggest challenge. The author
developed an expanded fog computing model with the integration of blockchain to share sensor data
among fog computing nodes. The new model is known as HCL-BaFog (Hypriot Cluster Lab) to collect
data and provide secure sharing of information to other fog nodes. The proof of concept that usages the
total virtualize function of the blockchain platform to ensure fairness in data sharing and the round-
robin block creation schedule method. The feasibility of the data-sharing model is tested by a testbed
of Raspberry Pi SBC (single board computer). This model uses the multichain framework to enhance
the performance of fog computing nodes and enable trust among peer nodes. However, this proposed
scheme is platform-dependent on the behavior of blockchain.
In (Lu et al., 2019) author address, the potential threat of data sharing that is data leakage and the
security of the network. given the author of the above issue presented the blockchain-based secure file
sharing model among different parties, Then introduced the information-sharing problem at machine
learning problems by bringing privacy secured federated learning. To maintain privacy this model uses
sharing model of machine learning instead of actual data. The blockchain-based collaborative architecture
is used to reduce the risk of data leakage in sharing the data among multiple entities. Federated learning
uses the concept of differential privacy to more strengthen data security. The normalized weighted graph
is used to generate the structure data instead of unstructured. The proof of training quality algorithm
converts the data-sharing problem in model sharing, this protects the privacy of the data owner. The
purpose of federated learning provides training to a data model that may provide correct responses for
information sharing request entities. This federated model used various machine learning algorithms
such as a random tree, random forest.
In (Manzoor et al., 2019) author discussed the issues of cloud-based data sharing centralize system
which requires third-party services to pay some fee for services. To overcome these scalability and trust
issues blockchain technology is used as a solution in IoT data sharing that uses the proxy re-encryption
technique to facilitate secured data transmission and visibility only to the data owner and intended user
into the blockchain network. The data is stored on the distributed cloud after passing through the en-
cryption process. The secure sharing is performed by a smart contract between the data owner and the
data requester. The whole framework is implemented on the ethereum virtualize platform to increase
the performance, reliability as well security features. The proposed architecture consists of four entity
includes IoT devices, data users, cloud-enabled service providers, and blockchain platforms. From the
security point of view, the author applied a certificate-based proxy re-encryption (CP-PRE) scheme that
consists of seven polynomial-time algorithms. The feasibility of the model is shown on permissioned
ethereum blockchain with the sensor of devices and cloud server for storing data. However this system is
not scalable to distributed cloud storage, so this will be the future enhancement of this proposed system.
In (Liang et al., 2019) author addresses the issues of previous data transmission techniques such as
low security, the higher management cost of data, lack of proper monitoring in the industrial Internet
of Things. These concerns may lead to tampering of data, unauthorized access in IoT devices. This
decreases the quality, consistency, and efficiency of quality data. In the view of above issues, the author
presented a secure blockchain-enabled data transmission model for the Industrial Internet of Things,
which using the dynamic secret sharing technique to provide secure transmission. The dynamic secret
sharing protects the private key. Key-value of private and public key sign the intelligent data transaction
210
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
as well as prove its validity of own transaction. In this paper author secured the data transmission by
blockchain sharing model that protects the decentralized system from attack. This model used the docker
virtualization technique for creating the power blockchain network sharing model. The experiment shows
that this model has achieved high security and efficient reliability, improves the transmission rate and
packet receiving rate. However this model needs a lot of enhancement in the future to introduce power
blockchain in large scalable networks, needs to remove data redundancies in the file storage system.
The paper (Stergiou et al., 2020) is related to provide the security and management of big data in fog
based environment for 6G wireless networks. This work provides the secure and efficient platform for
using internet and sharing information as well as provide scalability for large big data.
211
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
212
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
• Performance and scalability: These are the significant feature of blockchain in terms of process-
ing information. The performance depends upon the task the nodes perform like authentication,
verification, running the consensus mechanism, and maintained the synchronized copy of the
transaction to the global digital ledger (Croman et al., 2016). Another concern of blockchain is
scalability when several users increase in the network the computational overhead also increased
(Kosba et al., 2016).
• Storage capacity and privacy: As the blockchain is based on the decentralization concept to pro-
vide secure storage and access to transactions or data, store information on the ledger and main-
tain privacy among users for various services is the biggest challenge. The concern of decentral-
ized systems is leakage of the public and private key used for encryption and decryption, limited
storage capacity such as IoT devices (Xie et al., 2019). This needs to be improved in future work.
• Energy consumption: energy consumption is also associated with the above-mentioned issues.
It may happen when more users are connected to network or IoT environment in solving puzzles
the consensus algorithm like PoW requires hard computational resources in the mining process
to overcome these energy constraints many researchers presented the energy-efficient algorithm
such as proof of stake (PoS), delegated proof of stake (DpoS), practical byzantine fault tolerance
(PBFT), etc.(Xie et al., 2019).
CONCLUSION
Recently blockchain has gained a lot of recognition because of its features such as transparency, im-
mutability, decentralized environment, peer-to-peer node communication, and distributed ledger. It can
be used in diverse fields like the internet of things, healthcare, cloud storage, supply chain, smart cities,
etc. It is accompanied by transparency to the ledger’s data and transactions. In this survey, we have sum-
marized the existing work based on data sharing techniques using blockchain platforms in areas of IoT,
213
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
healthcare, and the internet of vehicles. Most of the authors proposed the security model that provides an
incentive mechanism for secure data sharing by using different Cryptographic algorithm and achieved the
protection of data up to some extent of security goal. However, there is still some hidden security threat
and challenges that need to improve more accurately in future work like scalability, energy consumption,
privacy and security of decentralize storage as well as integration of blockchain for fog computing to
provide better security, data integrity at fog layer, as fog layer help to provide better services with low
latency and to utilize full bandwidth.
REFERENCES
Amofa, S., Sifah, E. B., Kwame, O. B., Abla, S., Xia, Q., Gee, J. C., & Gao, J. (2018, September). A
blockchain-based architecture framework for secure sharing of personal health data. In 2018 IEEE 20th
International Conference on e-Health Networking, Applications and Services (Healthcom) (pp. 1-6).
IEEE. 10.1109/HealthCom.2018.8531160
Bhaskaran, K., Ilfrich, P., Liffman, D., Vecchiola, C., Jayachandran, P., Kumar, A., ... Teo, E. G. (2018,
April). Double-blind consent-driven data sharing on blockchain. In 2018 IEEE International Conference
on Cloud Engineering (IC2E) (pp. 385-391). IEEE. 10.1109/IC2E.2018.00073
Cash, M., & Bassiouni, M. (2018, September). Two-tier permission-ed and permission-less blockchain
for secure data sharing. In 2018 IEEE International Conference on Smart Cloud (SmartCloud) (pp.
138-144). IEEE. 10.1109/SmartCloud.2018.00031
Cech, H. L., Großmann, M., & Krieger, U. R. (2019, June). A fog computing architecture to share sensor
data by means of blockchain functionality. In 2019 IEEE International Conference on Fog Computing
(ICFC) (pp. 31-40). IEEE. 10.1109/ICFC.2019.00013
Croman, K., Decker, C., Eyal, I., Gencer, A. E., Juels, A., Kosba, A., & Song, D. (2016, February).
On scaling decentralized blockchains. In International conference on financial cryptography and data
security (pp. 106-125). Springer.
Dib, O., Brousmiche, K. L., Durand, A., Thea, E., & Hamida, E. B. (2018). Consortium blockchains:
Overview, applications and challenges. International Journal on Advances in Telecommunications, 11(1-2).
Dinh, T. T. A., Wang, J., Chen, G., Liu, R., Ooi, B. C., & Tan, K. L. (2017, May). Blockbench: A frame-
work for analyzing private blockchains. In Proceedings of the 2017 ACM International Conference on
Management of Data (pp. 1085-1100). 10.1145/3035918.3064033
Dubovitskaya, A., Novotny, P., Xu, Z., & Wang, F. (2020). Applications of blockchain technology
for data-sharing in oncology: Results from a systematic literature review. Oncology, 98(6), 403–411.
doi:10.1159/000504325 PMID:31794967
Eltayieb, N., Elhabob, R., Hassan, A., & Li, F. (2020). A blockchain-based attribute-based signcryption
scheme to secure data sharing in the cloud. Journal of Systems Architecture, 102, 101653. doi:10.1016/j.
sysarc.2019.101653
214
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
Feng, Q., He, D., Zeadally, S., Khan, M. K., & Kumar, N. (2019). A survey on privacy protection in block-
chain system. Journal of Network and Computer Applications, 126, 45–58. doi:10.1016/j.jnca.2018.10.020
Gai, K., Wu, Y., Zhu, L., Qiu, M., & Shen, M. (2019). Privacy-preserving energy trading using consortium
blockchain in smart grid. IEEE Transactions on Industrial Informatics, 15(6), 3548–3558. doi:10.1109/
TII.2019.2893433
Goyal, S. (2018). The History of Blockchain Technology: Must Know Timeline. Academic Press.
Guo, R., Shi, H., Zhao, Q., & Zheng, D. (2018). Secure attribute-based signature scheme with multiple
authorities for blockchain in electronic health records systems. IEEE Access: Practical Innovations,
Open Solutions, 6, 11676–11686. doi:10.1109/ACCESS.2018.2801266
Gupta, B. B., Li, K. C., Leung, V. C., Psannis, K. E., & Yamaguchi, S. (2021). Blockchain-assisted
secure fine-grained searchable encryption for a cloud-based healthcare cyber-physical system. IEEE/
CAA Journal of Automatica Sinica.
Gupta, S., & Gupta, B. B. (2017). Detection, avoidance, and attack pattern mechanisms in modern web
application vulnerabilities: Present and future challenges. International Journal of Cloud Applications
and Computing, 7(3), 1–43. doi:10.4018/IJCAC.2017070101
Hasselgren, A., Kralevska, K., Gligoroski, D., Pedersen, S. A., & Faxvaag, A. (2019). Blockchain in
healthcare and health sciences–a scoping review. International Journal of Medical Informatics, 104040.
PMID:31865055
Hofman, W. J. (2019). A Methodological Approach for Development and Deployment of Data Sharing
in Complex Organizational Supply and Logistics Networks with Blockchain Technology. IFAC-Paper-
sOnLine, 52(3), 55–60. doi:10.1016/j.ifacol.2019.06.010
Kosba, A., Miller, A., Shi, E., Wen, Z., & Papamanthou, C. (2016, May). Hawk: The blockchain model
of cryptography and privacy-preserving smart contracts. In 2016 IEEE symposium on security and
privacy (SP) (pp. 839-858). IEEE.
Liang, W., Tang, M., Long, J., Peng, X., Xu, J., & Li, K. C. (2019). A secure fabric blockchain-based data
transmission technique for industrial Internet-of-Things. IEEE Transactions on Industrial Informatics,
15(6), 3582–3592. doi:10.1109/TII.2019.2907092
Liu, C. H., Lin, Q., & Wen, S. (2018). Blockchain-enabled data collection and sharing for industrial
IoT with deep reinforcement learning. IEEE Transactions on Industrial Informatics, 15(6), 3516–3526.
doi:10.1109/TII.2018.2890203
Liu, X., Wang, Z., Jin, C., Li, F., & Li, G. (2019). A Blockchain-Based Medical Data Sharing and Pro-
tection Scheme. IEEE Access: Practical Innovations, Open Solutions, 7, 118943–118953. doi:10.1109/
ACCESS.2019.2937685
Liu, Y., Zhang, J., & Gao, Q. (2018, October). A Blockchain-Based Secure Cloud Files Sharing Scheme
with Fine-Grained Access Control. In 2018 International Conference on Networking and Network Ap-
plications (NaNA) (pp. 277-283). IEEE. 10.1109/NANA.2018.8648778
215
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
Lu, Y., Huang, X., Dai, Y., Maharjan, S., & Zhang, Y. (2019). Blockchain and Federated Learning for
Privacy-preserved Data Sharing in Industrial IoT. IEEE Transactions on Industrial Informatics.
Manzoor, A., Liyanage, M., Braeke, A., Kanhere, S. S., & Ylianttila, M. (2019, May). Blockchain based
proxy re-encryption scheme for secure IoT data sharing. In 2019 IEEE International Conference on
Blockchain and Cryptocurrency (ICBC) (pp. 99-103). IEEE. 10.1109/BLOC.2019.8751336
McGhin, T., Choo, K. K. R., Liu, C. Z., & He, D. (2019). Blockchain in healthcare applications: Research
challenges and opportunities. Journal of Network and Computer Applications, 135, 62–75. doi:10.1016/j.
jnca.2019.02.027
Mohan, A. P., & Gladston, A. (2020). Merkle tree and Blockchain-based cloud data auditing. Interna-
tional Journal of Cloud Applications and Computing, 10(3), 54–66. doi:10.4018/IJCAC.2020070103
Mohanta, B. K., Jena, D., Panda, S. S., & Sobhanayak, S. (2019). Blockchain Technology: A Survey on
Applications and Security Privacy Challenges. Internet of Things, 100107.
Nakasumi, M. (2017, July). Information sharing for supply chain management based on block chain
technology. In 2017 IEEE 19th Conference on Business Informatics (CBI) (Vol. 1, pp. 140-149). IEEE.
10.1109/CBI.2017.56
Naz, M., Al-zahrani, F. A., Khalid, R., Javaid, N., Qamar, A. M., Afzal, M. K., & Shafiq, M. (2019). A
Secure Data Sharing Platform Using Blockchain and Interplanetary File System. Sustainability, 11(24),
7054. doi:10.3390u11247054
Nguyen, D. C., Pathirana, P. N., Ding, M., & Seneviratne, A. (2019). Blockchain for secure EHRs shar-
ing of mobile cloud based e-Health systems. IEEE Access: Practical Innovations, Open Solutions, 7,
66792–66806. doi:10.1109/ACCESS.2019.2917555
Pham, H. A., Le, T. K., & Le, T. V. (2019, September). Enhanced Security of IoT Data Sharing Manage-
ment by Smart Contracts and Blockchain. In 2019 19th International Symposium on Communications
and Information Technologies (ISCIT) (pp. 398-403). IEEE. 10.1109/ISCIT.2019.8905219
Rahman, M. S., Al Omar, A., Bhuiyan, M. Z. A., Basu, A., Kiyomoto, S., & Wang, G. (2020). Account-
able cross-border data sharing using blockchain under relaxed trust assumption. IEEE Transactions on
Engineering Management.
Rawat, D. B., Njilla, L., Kwiat, K., & Kamhoua, C. (2018, March). iShare: Blockchain-based privacy-
aware multi-agent information sharing games for cybersecurity. In 2018 International Conference on
Computing, Networking and Communications (ICNC) (pp. 425-431). IEEE. 10.1109/ICCNC.2018.8390264
Samuel, O., Javaid, N., Awais, M., Ahmed, Z., Imran, M., & Guizani, M. (2019, July). A blockchain
model for fair data sharing in deregulated smart grids. In IEEE Global Communications Conference
(GLOBCOM 2019). 10.1109/GLOBECOM38437.2019.9013372
Shrestha, A. K., Deters, R., & Vassileva, J. (2019). User-controlled privacy-preserving user profile data
sharing based on blockchain. arXiv preprint arXiv:1909.05028.
216
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
Shrestha, A. K., & Vassileva, J. (2018, June). Blockchain-based research data sharing framework for
incentivizing the data owners. In International Conference on Blockchain (pp. 259-266). Springer.
10.1007/978-3-319-94478-4_19
Si, H., Sun, C., Li, Y., Qiao, H., & Shi, L. (2019). IoT information sharing security mechanism based
on blockchain technology. Future Generation Computer Systems, 101, 1028–1040. doi:10.1016/j.fu-
ture.2019.07.036
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based big data secure management in the
fog over a 6G wireless network. IEEE Internet of Things Journal, 8(7), 5164–5171. doi:10.1109/
JIOT.2020.3033131
Sumathi, M., & Sangeetha, S. (2020). Blockchain based sensitive attribute storage and access moni-
toring in banking system. International Journal of Cloud Applications and Computing, 10(2), 77–92.
doi:10.4018/IJCAC.2020040105
Theodouli, A., Arakliotis, S., Moschou, K., Votis, K., & Tzovaras, D. (2018, August). On the design
of a Blockchain-based system to facilitate Healthcare Data Sharing. In 2018 17th IEEE International
Conference on Trust, Security And Privacy In Computing And Communications/12th IEEE International
Conference On Big Data Science And Engineering (TrustCom/BigDataSE) (pp. 1374-1379). IEEE.
10.1109/TrustCom/BigDataSE.2018.00190
Thwin, T. T., & Vasupongayya, S. (2018, August). Blockchain based secret-data sharing model for per-
sonal health record system. In 2018 5th International Conference on Advanced Informatics: Concept
Theory and Applications (ICAICTA) (pp. 196-201). IEEE. 10.1109/ICAICTA.2018.8541296
Wang, H., & Song, Y. (2018). Secure cloud-based EHR system using attribute-based cryptosystem and
blockchain. Journal of Medical Systems, 42(8), 152. doi:10.100710916-018-0994-6 PMID:29974270
Wang, H., Zhang, Y., Chen, K., Sui, G., Zhao, Y., & Huang, X. (2019). Functional broadcast encryption
with applications to data sharing for cloud storage. Information Sciences, 502, 109–124. doi:10.1016/j.
ins.2019.06.028
Wang, L., Liu, W., & Han, X. (2017, December). Blockchain-based government information resource
sharing. In 2017 IEEE 23rd International Conference on Parallel and Distributed Systems (ICPADS)
(pp. 804-809). IEEE. 10.1109/ICPADS.2017.00112
Wang, Y., Zhang, A., Zhang, P., & Wang, H. (2019). Cloud-Assisted EHR Sharing With Security and
Privacy Preservation via Consortium Blockchain. IEEE Access: Practical Innovations, Open Solutions,
7, 136704–136719. doi:10.1109/ACCESS.2019.2943153
Wang, Z., Tian, Y., & Zhu, J. (2018, August). Data sharing and tracing scheme based on blockchain.
In 2018 8th International Conference on Logistics, Informatics and Service Sciences (LISS) (pp. 1-6).
IEEE. 10.1109/LISS.2018.8593225
Wang, Z., & Wu, Q. (2019, October). Incentive for Historical Block Data Sharing in Blockchain. In
2019 IEEE 10th Annual Information Technology, Electronics and Mobile Communication Conference
(IEMCON) (pp. 0913-0919). IEEE. 10.1109/IEMCON.2019.8936209
217
A Study on Data Sharing Using Blockchain System and Its Challenges and Applications
Wu, A., Zhang, Y., Zheng, X., Guo, R., Zhao, Q., & Zheng, D. (2019). Efficient and privacy-preserving
traceable attribute-based encryption in blockchain. Annales des Télécommunications, 74(7-8), 401–411.
doi:10.100712243-018-00699-y
Xie, J., Tang, H., Huang, T., Yu, F. R., Xie, R., Liu, J., & Liu, Y. (2019). A survey of blockchain tech-
nology applied to smart cities: Research issues and challenges. IEEE Communications Surveys and
Tutorials, 21(3), 2794–2830. doi:10.1109/COMST.2019.2899617
Xu, L., Shah, N., Chen, L., Diallo, N., Gao, Z., Lu, Y., & Shi, W. (2017, April). Enabling the sharing
economy: Privacy respecting contract based on public blockchain. In Proceedings of the ACM Workshop
on Blockchain, Cryptocurrencies and Contracts (pp. 15-21). 10.1145/3055518.3055527
Xuan, S., Zheng, L., Chung, I., Wang, W., Man, D., Du, X., & Guizani, M. (2020). An incentive mecha-
nism for data sharing based on blockchain with smart contracts. Computers & Electrical Engineering,
83, 106587. doi:10.1016/j.compeleceng.2020.106587
Zhang, Y., He, D., & Choo, K. K. R. (2018). BaDS: Blockchain-based architecture for data shar-
ing with ABS and CP-ABE in IoT. Wireless Communications and Mobile Computing, 2018, 2018.
doi:10.1155/2018/2783658
218
219
Chapter 12
Fruit Fly Optimization-
Based Adversarial Modeling
for Securing Wireless
Sensor Networks (WSN)
Priyanka Ahlawat
National Institute of Technology, Kurukshetra, India
Mukul Goyal
National Institute of Technology, Kurukshetra, India
Rishabh Sethi
National Institute of Technology, Kurukshetra, India
Nitish Gupta
National Institute of Technology, Kurukshetra, India
ABSTRACT
Node capture attack is one of the crucial attacks in wireless sensor networks (WSN) that seizes the
node physically and withdraws the confidential data from the node’s memory. The chapter exploits the
adversarial behavior during a node capture to build an attack model. The authors also propose a fruit
fly optimization algorithm (FFOA) that is a multi-objective optimization algorithm that consists of a
number of objectives for capturing a node in the network: maximum node contribution, maximum key
contributions are some examples of the same. The aim is to demolish the maximum part of the network
while minimizing the cost and maximizing attacking efficiency. Due to the multi-objective function, the
authors attain a maximum fraction of compromised traffic, lower attacking rounds, and lower energy
cost as contrasted with other node capture attack algorithms. They have developed an algorithm, which
is an enhanced version of FFOA and has even better efficiency than FFOA.
DOI: 10.4018/978-1-7998-7789-9.ch012
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
INTRODUCTION
Node capture attack is a comprehensive attack in which the intruder physically captures the sensor
node by extracting keys and confidential data. With technological advances in the field of wireless sen-
sor technology, various operations such as catastrophic and defense monitoring can be painlessly and
quickly deployed to wireless sensor networks (WSN)(Lin, C,2016) . The scattered nodes communicate
wirelessly to a central gateway, which connects to the wired world where users can collect, process,
analyze, and present the measured physical data. Though WSNs have their advantages as the nodes are
autonomous, they still need to be addressed. Some of the most common and important challenges are
coverage, scalability, QoS, and security. Amidst the challenges listed, security is a major issue to be
addressed in WSNs. Sensor networks is extremely vulnerable to node capture attacks. WSN is a group
of a huge number of low price, low control, and self-organizing specialized sensor nodes (Lin, C,2015).
It is very much vulnerable to different physical attacks due to limited resource capacity and screened
to the external atmosphere for circulating network data. The node capture attack is one of the major at-
tacks in WSN in which the attacker physically captures the node and can remove the secret information
from the node’s memory or misuse the confidential data (Lin.2013). With technological advances in
wireless sensor technology, various operations such as the health and defense monitoring can be quickly
deployed to WSN. We focus on developing a multi-objective function using which we can compromise
the network efficiently and quickly, unlike random attack (RA), maximum key attack (MKA), maximum
link attack (MLA). WSN is a wireless network comprising a large number of self-operative and self-
sufficient nodes, which comprises low cost, less control, and self-organizing qualities(Lin.2013). This
type of network uses sensors for catastrophic and defense monitoring of physical and environmental
situations. When such self-governing nodes are used with routers and gateways, it creates a wireless
sensor network system. This technology has its applications in various fields, including military, medi-
cal, defense, environmental, and many more.There are various issues to be addressed in WSNs. Some
of them are scalability, quality of service, size, security, and many more (Tague, P,2008). Out of all
these, security is the biggest challenge. Due to tight resource capacity and its exposure to the outer en-
vironment, it is prone to various kinds of physical attacks. Broadly, attacks can be of two types: active
attacks, including routing attacks, eavesdropping, and passive attacks, which include all attacks against
privacy. The performance is measure with other methods, and thus gives the improved resilience in
order capturing node, hash computations decreased, compromise probability for proxy nodes also re-
duced with a revoked link (Ahlawat,2018) . The result matrix is examined with old strategies in order
of the number attacking rounds, capturing cost and traffic compromised.. The performance validated by
number of path compromise, path length, and route ratio (Ahlawat,2018). With this analysis of keys and
linear automated theory, develop a model that effectively describes the behaviour of that network with
attack. Optimal control theory method design a response for the network, which provide a network with
secure stability(Bonaci,2010).Node capture attacks are one of the most major attacks in WSN. Node
Capture Attack is a kind of attack in which the intruder can access the entire network and perform any
operation on the network. The attacker captures the sensor node by gaining access to cryptographic keys
and secret information like key pre-distribution model(Shukla,2015). Earlier, node capture attacks were
having limitations like a lack of attacking methods and low attack efficiency. There are different types
of node capture attacks. The next stage of the internet’s development i.e. the internet of things which
makes the internet a physical network requires the objects to communicate with minimal human interfer-
ence. This type of network made of mobile sensor nodes communicating with each other and working
220
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
synchronously in a controlled envirnments .It is a wireless network that comprises a great number of
dynamic, self-directed, small, self operative, low powered devices named sensor nodes called ‘motes’.
A large number of battery-operated, spatially separated and minute devices are networked to collect,
process, and transfer data to the operators in these networks. It has also controlled the ability to compute
and to process the received data. It finds its application in numerous fields such as - Home Applications,
Commercial Applications, Forest fire detection, Area monitoring, Air pollution monitoring, Military
Applications, Health care monitoring, Health Applications, Environmental Applications, Earth/Environ-
mental sensing, Water quality monitoring, Landslide detection, Industrial monitoring, and many more
applications. Thus, so many applications of WSN make its security a prime concern. Henceforth, it aims
to find out the vulnerability and weak points in WSN by simulating multiple attacks and concludes to
point out where a WSN must be made more robust. Thus, to make the next generation of the internet,
i.e., IoT, more reliable and secured, it is important to find out the most probable and destructive attacks
prone to affect WSN’s security. To conclude, we have simulated the most advanced attack, i.e., FFOA
on node capture attack (Bhatt,2020). Further, a comparative study is provided by plotting graphs for
FFOA and its improved version. Therefore, to enhance the attack efficiency of node capture attack, we
propose an attacking approach using the Fruit Fly Optimization Algorithm (FFOA), which overcomes
these limitations. We aim to develop a multi-objective function using which we can compromise the
network efficiently and quickly. We would then compare the performance with few similar competing
strategies to show that the FFOA outperforms its competitor.
MOTIVATION
The next stage of the internet’s development i.e. the Internet of things which makes the internet a physical
network requires the objects to communicate with minimal human interference. This type of network made
of mobile sensor nodes communicating with each other and working synchronously is known as WSN.
As it is a wireless network that comprises a great number of dynamic, self-directed, small, low powered
devices named sensor nodes called motes. A huge number of battery-operated, spatially separated and
minute devices are networked to collect, process, and transfer data to the operators in these networks.
It also has controlled the ability to compute and to process the received data. It finds its application
in numerous fields such as - Home Applications, Commercial Applications, Air pollution monitoring,
Forest fire detection, Area monitoring, Military Applications, Health Applications, Environmental Ap-
plications, Health care monitoring, Environmental/Earth sensing, Landslide detection, Water quality
monitoring, Industrial monitoring. Thus, such a vast application of WSN makes its security a prime
concern. Hence, we aim to find out the vulnerability and weak points in a WSN by simulating multiple
attacks and plotting a comparison graph to point out the fields or parts where a WSN must be made
more secure. Thus, to make the next generation of internet i.e. IoT more trustworthy and secure we aim
to find out the most probable and devastating attack which can affect WSN seriously. To achieve this,
we simulate various attacks such as Random attack, Maximum Key Attack, Maximum Link attack, and
applying FFOA on node capture attack. Further, we provide a comparative study by plotting graphs for
each of the above attacks.
221
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
LITERATURE SURVEY
This section details the various schemes proposed to increase the attacking efficiency of attacker. WSN
is a wireless network, comprising of large number of self-governing nodes, which are of low cost, less
control and are self-organizing. This type of network uses sensors for catastrophic and defense moni-
toring of physical and environmental situations. When such self-governing nodes are used with routers
and gateways, it creates a Wireless sensor network system. Currently, these types of network are in the
initial stage of being deployed. But, in next few years, one would find these networks throughout the
world, having access to them via Internet. This technology has its applications in various fields including
military, medical, defense, environmental and many more.
Node capture is a kind of attack in which the intruder can gain the access over entire network and
perform any operation on the network. In this, the attacker physically captures the sensor node by gaining
access to cryptographic keys and secret information. Earlier, node capture attacks were having limita-
tions like lack of attacking methods and low attack efficiency. There are different types of node capture
attacks which are given below:
222
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
The attacks mentioned above were not efficient enough and they had some major issues, due to which
the attack efficiency of these attacks was low. Some of these issues are addresses below.
1. Basically, susceptibility is defined as a number, and it cannot specifically describe the level of
destruction in wireless sensor networks.
2. The attack efficiency of node capture attack is minimum, because it desires to capture the maximum
amount of nodes in network.
3. The domain of many of these approaches is limited and is therefore not suitable for providing ap-
propriate results.
4. The most important problem is that most of these methods focused only on single criteria like
lowest cost, maximum keys or any such criteria.
So, to improve the attack efficiency in node capture attack, an approach should be developed which
focus on multiple objectives, like lowest cost, maximum key contribution, degree of nodes, and similarly,
we can include many more. This method can be used to find out the optimal nodes using optimization
algorithms (Genetic algorithm or Fruit Fly Optimization algorithm). Several node capture attack algo-
rithms have been given by the researchers to increase the capability of attacker in WSN. This kind of
attack models can be exploited by the network designers to improve the security of complete network.
The node capture attack is one of the foremost attacks in WSN. In this attack, the intruder physically
captures the sensor node and eradicates the secret keying information from the node’s memory. Several
optimizations algorithms are also applied to increase the capability of adversary. A Fruit Fly Optimiza-
tion Algorithm (FFOA) is used to increase the attacker capacity which consists of several objectives
223
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
namely maximum node contribution, maximum node key contribution, least resource expenses, minimum
energy cost to determine the set of optimal nodes. These optimal nodes are the nodes to be captured to
destroy the complete network. It will result in destroying the maximum part of the network with mini-
mum energy cost and maximum attacking efficiency. The performance parameters are compromised
traffic, number of attacking rounds, and energy cost. The attacking models given above are still not very
efficient, due to which their attacking efficiency is very low. The following issues need to be addressed
during a design of attacking model:
• To overcome the problem of vulnerability based approaches and thus, to improve the attacking
effectiveness or efficiency of the node capture attacking model.
• To design an effective model based on optimization algorithm that may have multiple objectives
such as least number of sensor nodes, highest node key contribution, and minimum number of
network resources, energy cost to find out an optimal set of node using Optimization Algorithms
(like Fruit Fly Optimization).
• To construct algorithm that may cause higher destructiveness in the WSN with least number of
resource expenses.
Following are the models that are used to design an effective node capture attack:
1. Network Model
WSN is represented by the directed network graph G = (N, L), where N is the nodes number and
L is the links number.
2. Key Distribution Model
The cryptographic keys represent a key group set K and every sensor node Nx∈N is randomly
assigned a subset of keys Kx⊂K from a key group set. Two sensor nodes Nx and Ny share a set of
cryptographic keys Kx,y= Kx∩Ky. These nodes are located in each other transmission range r and
can communicate with each other. The set of Kx,y encrypt the messages transferred between Nx and
Ny. For example: Kx= {K1,K5,K8,K4} and Kb= {K3,K5,K7,K8}, then Kx,y= KxÇKy ={K5,K8}
3. Link Model
A link Lx,y is reliable and confined if it is encrypted by Kx,y. The information or messages can thus
be sent between Nx and Ny without revealing it to other nodes. The security strength of the link Lx,y
is directly dependent on the size of Kx,y. The larger the overlapping value, higher is the strength.
We specify the set of all the links in WSN by L, where L= {Lx,y| Nx∈N, Ny∈N}.
4. Adversary Model
This model is described from the view of an attacker’s and it is theoretical that the intruder can get
the access to the information transmitting through the WSN. The intruder’s focus is to do attack
to capture the complete network by achieving more than one objective. It results in lower resource
expenses, a large number of captured keys and least number of nodes.
To discover the optimal nodes to be captured, an optimization algorithm must be used. There are
various optimization algorithms available. The Fruit Fly Optimization Algorithm (FOA) is a bio inspired
algorithm for discovering global optimization based on the food searching performance of the fruit fly.
The fruit fly is better than other species in sensing and perception, particularly, in vision. The fruit fly
flies from one fruit to another using its sensitive visualization to discover the fruit which appears to be
224
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
the best at the moment and then it flies in that path. FFOA works in a similar manner by computing
the multi objective function, this algorithm is initiated to learn optimal nodes from the existing set of
node, which minimize the multi objective function to produce the optimal results. We know that there
are different methods to perform node capture attack in WSN. Our aim to is to enhance the attack ef-
ficiency of node capture attack in these networks. Different approaches have different efficiencies in
terms of number of attacking round, cost required in compromising the network, and other parameters
too. So, we propose to find out the best approach which would compromise the complete network in
minimum attacking rounds and least cost. Different routing paths that contain multiple paths, are used
for compromising the complete network traffic. The following matrices are used to find out the most
destructiveness sensor node.
• Key Route Matrix (KR): This matrix determines the relationship between the cryptographic
keys that are stored in every sensor node and the different routes that connect source node to the
destination node.
• Vertex (Sensor Node)-Key Matrix (VK): This matrix represents the relationship between the
keys that are stored in every sensor node. It is the number of keys stored in sensor node.
• Key-Number Matrix (KN): This matrix represent the relationship between the cryptographic
keys that are associated with every sensor node..
• VLR: To assess on the partial keying relationship between nodes and the routes, we calculate
another matrix VLR. It is used to demonstrate the ratio of overlapping keys captured by the adver-
sary during a node capture attack.
• Vertex (Sensor Node)- Route Matrix(VR): This matrix gives the relationship between the sen-
sor node and the different routes that connect source node to the destination node.
Where VR = VK * KR.
1 if Ka can coperate Kb
KRa ,b (1)
0 otherwise
1 if Ka b
VK b ,a (2)
0 otherwise
k
KN b
VK b ,a if Ka b
a 1 (3)
0
otherwise
VR = VK * KR (4)
225
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
R
MM b
MM b ,a participation of node Nb
a 1 (6)
0
otherwise
CSb ,a
CSb ,a = (7)
Wb
R
CSb
CSb ,a Cost node Nb
a 1 (8)
0
otherwisee
N R
1 1
Fb CSb (9)
b 1 a 1 MMb KNb
To discover the optimal nodes to be captured, an optimization algorithm must be used. There are
various optimization algorithms available. Here, one of them is discussed.
226
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
The Fruit Fly Optimization Algorithm (FOA) is a bio-inspired algorithm for finding global opti-
mization based on the food searching behavior of the fruit fly. The fruit fly is better than other species
in sensing and perception, particularly, in vision. The fruit fly flies from one fruit to another using its
sensitive visualization to discover the fruit which appears to be the best at the moment and then it flies
in that path. FFOA works in a similar manner by estimating multi objective function, this algorithm is
used to discover optimal nodes from the existing node, which minimize the multi objective function to
produce the optimal results (Bhatt,2020).
Different techniques for node capture attack have been discussed. Fruit Fly Optimization Algorithm
has also been discussed here. They have shown that using Fruit Fly Optimization algorithm, the attack
efficiency can be increased. Node capture techniques are classified as centralized and distributed schemes.
MA, FGA, MREA are examples of distributed attack whereas random key predistribution is an example
of centralized attack (Changlani et al., 2020).
Authors have given a matrix approach for implementing node capture attack. They have shown that
Matrix approach is better than random attack, maximum link attack and maximum node attack (Lin &
Wu 2013).
In this technique, various matrices are created to store the vulnerability value of every node of the
network. The node with highest value of destructiveness is selected in each round. After every round, the
values of matrices are modified. It includes a high complexity. MA has lesser number of captured nodes
thus least energy cost. However MA has not well addressed the energy cost and attacking efficiency(Lin,
C,2016)
Authors have discussed about node capture attack and have given different approaches to detect the
node capture attack in WSN. But these approaches are also not efficient, some more methods need to
be developed (Selvamani et al., 2012)
Authors have used flow of current from an electric circuit to represent a compromise traffic. It is used
to quantify the vulnerability of network. A metric named vulnerability is given and it is calculated based
on routing path. A minimum cost node capture attack model is given that is based on greedy heuristic.
It results in lesser energy cost and maximum destructiveness in the network (Poovendran et al., 2008).
Authors have discussed about FFOA and its improved versions and hybrid FOA approaches. We
know that there are different methods to perform node capture attack in WSN. Our aim to is to enhance
the attack efficiency of node capture attack in these networks. Different approaches have different ef-
ficiencies in terms of number of attacking round, cost required in compromising the network, and other
parameters too. So, we propose to find out the best approach which would compromise the complete
network in minimum attacking rounds and least cost. For this purpose, we would create a random net-
work and simulate different techniques of node capture attack on it. Then, we would do a comparative
study of the results generated by these attacks. But as theoretical studies propose, these attacks have
several drawbacks. So, we would run optimization algorithms on top of these attacks, to generate op-
timal results. Firstly, we have performed three basic attacks on our network, namely, Random Attack,
Maximum Key Attack and Maximum Link Attack. We have provided the pseudo codes for these attack
algorithms (Iscan & Gunduz, 2015).
WSNs are generally placed in hostile areas thus node capture is one of the critical attack. an adversarial
model is given based on different vulnerability factors of the network such as node density, placement
of sink node, neighbor influence factor that enhances the capability of attacker (Dave et al., 2018).
227
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
In data aggregation, node capture can significantly reduce the performance . It exposes the network
to leakage of confidentiality. In this technique the cluster head selects the set of nodes. During data
transfer, portions of data are transferred encrypted by authenticated keys. When a node receives the data
portions then, it combines them and send to cluster head. Thus, it is able to mitigate the effect of node
capture impact on the network (Bhoopathy & Parvathi 2012).
Authors presented multiple objective such as maximum node compromise, key participation, resource
expenditure. It is used to find the optimal set of nodes to capture the complete network. PSO and GA are
used to find this optimal set. It is shown that FiRAO-PG has higher destructiveness than MA (Tantubay
et al., 2015).
Authors emphasized the maintenance of network connectivity in net centric warfare. In this paper
several factors that can influence the connectivity are studied such as key predistribution. A relationship
between network connectivity and key predistribution is given . An expression is formulated to find out
the required communication radius for such kind of applications (Chan & Fekri, 2007).
Pseudo Code 3: Maximum Link Attack (nodes, paths, VK, cost, adjList)
1. Initialize the arrays to store performance of attack.
2. While network is not compromised do
3. Select the node having maximum neighbors.
4. Attack the selected node.
5. Update the cost and keys acquired.
6. Compromise the network using the nodes captured.
7. End while
8. Return attack statistics.
228
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
After performing these three basic attacks, we have implemented Fruit Fly Optimization Algorithm,
which needs different matrices to be calculated upon which it performs optimization. These matrices
show the relation of nodes with keys and routes.
After performing these attacks, we have developed an algorithm which is an optimization over the
Fruit Fly Optimization Algorithm (FFOA), where, after capturing a node we are re-evaluating the ma-
trices over the network again by considering the already captured keys and not using them again for the
remaining un captured nodes. This way the results produced are more optimal and hence, the number
of rounds and the energy cost are minimized.
After implementing all these approaches, we have provided a comparative result in the form of dif-
ferent plots over different parameters.
229
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
1 Level 0 DFD
2 Level 1 DFD.
230
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
3 Level 2 DFD
IMPLEMENTATION DETAILS
Initially, we create a random network of predetermined number of nodes (here, we have taken number
of nodes to be 25). Then we allocate some number of keys randomly to each node (here, we have taken
a total of 500 keys, and to each node, we have randomly allocated 3 to 15 keys), and we are storing this
network using adjacency matrices. We have selected a node as the source vertex and a node as destina-
tion vertex. We have created all routes from source to destination. We have assumed that the attacker
has information about the key pre-distribution model and routing mechanism used in the WSN. Also,
we have allocated some predefined cost of capturing each vertex. Then, we have implemented 3 basic
attacks, namely, Random Attack, Maximum Key Attack and Maximum Link Attack. After these three
basic attacks, we have implemented fruit fly optimization algorithm, which requires some matrices to
be computed that store the relations between nodes, routes and keys. Finally, we have developed an ef-
ficient algorithm, which is an optimized version of FFOA.
By performing different attacks on the randomly created network, we have gathered sufficient informa-
tion about different parameters, by which we can compare the efficiency of these attacks.
If we notice the number of nodes to be captured to compromise the whole network, we observed that
the proposed algorithm outperforms the other algorithms, particularly in this aspect. It requires minimum
number of nodes to be captured to compromise the complete traffic. It is shown in Fig 6. It is due the
fact we have improved the basic FFOA to further minimize the number of captured nodes.
231
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
Figure 7 shows the relationship between the final cost and the number of keys captured.
We observed that energy cost is minimum in enhanced FFOA, as compared with the other attacks.
Also, the number of keys to be captured is least in this improved algorithm.
Figure 8, relation between the cost and number of captured nodes is shown and it is noticed that
the cost increases slowly in enhanced FFOA, and is also minimum, when compared to its competitors.
This chapter concludes that the enhanced node capture attack outperforms the basic attacks like MLA,
MKA and RA in almost all perspectives. The reason for such improved performance of FFOA as com-
pared to other algorithms is its consideration to multiple objectives at every point whereas the simpler
algorithms mostly work on a single objective which is either capturing node with maximum key, maxi-
232
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
mum links or minimum cost. But FFOA takes note of all these parameters and captures the node which
yields maximum damage in terms of links and keys with keeping minimum cost in mind. The improved
FFOA works even better due to its dynamic nature towards current network and hence predicting more
precisely the node which yields maximum damage. The multiple objective function in advanced al-
gorithms can also be adjusted in reference to contribution of keys and links and hence provides more
dynamic and adjusting way to perform an efficient attack. The main problem with advanced or improved
attacks is their time complexity with respect to number of nodes and paths. These algorithms take way
longer time as compared to basic algorithms and hence can be improved further by inculcating dynamic
behavior along with consideration to time taken. Though the results were promising there is still scope
of enhancement in future. Firstly, the time taken to attack a network is currently very large so we can
work on the time complexity aspect of the improved FFOA. The other thing we can work on is making
our attack model responsive to the changes in WSN. As WSN is a mobile network therefore the topology
and hence the connectivity of the network keeps changing with time as new nodes are added to network
and existing nodes leave network but our assumption of fixed nodes with fixed locations doesn’t allow us
to incorporate these changes. The other thing we can experiment upon is changing the criteria of selec-
tion of node. Currently the node to be attacked in improved FFOA is searched over the entire network
which can be changed to some subset of nodes instead of searching over the entire network. As a future
work, this algorithm can be applied to clustered senor networks and WSN with mobile nodes. Further,
optimization algorithms like simulated annealing, binary PSO, genetic algorithm may also be used to
minimize the number of captured nods of the network.
REFERENCES
Ahlawat, P., & Dave, M. (2018). An attack model based highly secure key management scheme for
wireless sensor networks. Procedia Computer Science, 125, 201–207. doi:10.1016/j.procs.2017.12.028
Ahlawat, P., & Dave, M. (2018). An attack resistant key predistribution scheme for wireless sensor
networks. Journal of King Saud University-Computer and Information Sciences.
Bharathi, Tanguturi, JayaKumar, & Selvamani. (2012). Node capture attack in Wireless Sensor Network:A
survey. 2012 IEEE International Conference on Computational Intelligence and Computing Research,
ICCIC 2012, 1-3. 10.1109/ICCIC.2012.6510237
Bhatt, R., Maheshwary, P., Shukla, P., Shukla, P., Shrivastava, M., & Changlani, S. (2020). Implementa-
tion of Fruit Fly Optimization Algorithm (FFOA) to escalate the attacking efficiency of node capture
attack in Wireless Sensor Networks (WSN). Computer Communications, 149, 134–145. doi:10.1016/j.
comcom.2019.09.007
Bonaci, T., Bushnell, L., & Poovendran, R. (2010, December). Node capture attacks in wireless sensor
networks: A system theoretic approach. In 49th IEEE Conference on Decision and Control (CDC) (pp.
6765-6772). IEEE. 10.1109/CDC.2010.5717499
Chan, H., Perrig, A., & Song, D. (2003, May). Random key predistribution schemes for sensor networks.
In 2003 Symposium on Security and Privacy, 2003 (pp. 197-213). IEEE. 10.1109/SECPRI.2003.1199337
233
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
Chan, K., & Fekri, F. (2007, May). Node compromise attacks and network connectivity. In Defense
Transformation and Net-Centric Systems 2007 (Vol. 6578, p. 65780W). International Society for Optics
and Photonics.
Chhabra, Gupta, & Almomani. (2013). A novel solution to handle DDOS attack in MANET. Academic
Press.
De, P., Liu, Y., & Das, S. K. (2006, June). Modeling node compromise spread in wireless sensor networks
using epidemic theory. In 2006 International Symposium on a World of Wireless, Mobile and Multimedia
Networks (WoWMoM’06). IEEE. 10.1109/WOWMOM.2006.74
Du, W., Deng, J., Han, Y. S., Chen, S., & Varshney, P. K. (2004, March). A key management scheme
for wireless sensor networks using deployment knowledge. In IEEE INFOCOM 2004 (Vol. 1). IEEE.
Ehdaie, M., Alexiou, N., Ahmadian, M., Aref, M. R., & Papadimitratos, P. (2017). Mitigating Node
Capture Attack in Random Key Distribution Schemes through Key Deletion. Journal of Communication
Engineering, 6(2), 99–109.
Eschenauer, L., & Gligor, V. D. (2002, November). A key-management scheme for distributed sensor
networks. In Proceedings of the 9th ACM conference on Computer and communications security (pp.
41-47). 10.1145/586110.586117
Iscan, H., & Gunduz, M. (2015, November). A survey on fruit fly optimization algorithm. In 2015 11th
International Conference on Signal-Image Technology & Internet-Based Systems (SITIS) (pp. 520-527).
10.1109/SITIS.2015.55
Jain, A. K., & Gupta, B. B. (2019). A machine learning based approach for phishing detection using
hyperlinks information. Journal of Ambient Intelligence and Humanized Computing, 10(5), 2015–2028.
doi:10.100712652-018-0798-z
Lin, C., Qiu, T., Obaidat, M. S., Yu, C. W., Yao, L., & Wu, G. (2016). MREA: A minimum resource
expenditure node capture attack in wireless sensor networks. Security and Communication Networks,
9(18), 5502–5517. doi:10.1002ec.1713
Lin, C., Wu, G., Yu, C. W., & Yao, L. (2015). Maximizing destructiveness of node capture attack in wire-
less sensor networks. The Journal of Supercomputing, 71(8), 3181–3212. doi:10.100711227-015-1435-7
Lin & Guowei Wu. (2013). Enhancing the attacking efficiency of the node captureattack in WSN: a
matrix approach. J Supercomput, Springer Science &Business Media, 1-19.
Qin, T., & Chen, H. (2012). An Enhanced Scheme against Node Capture Attack using Hash-Chain for Wire-
less Sensor Networks. Journal of Information Technology, 11(1), 102–109. doi:10.3923/itj.2012.102.109
Shaila, K., Manjula, S. H., Thriveni, J., Venugopal, K. R., & Patnaik, L. M. (2011). Resilience against
node capture attack using asymmetric matrices in key predistribution scheme in wireless sensor networks.
International Journal on Computer Science and Engineering, 3(10), 3490.
Shukla, P. K., Goyal, S., Wadhvani, R., Rizvi, M. A., Sharma, P., & Tantubay, N. (2015). Finding robust
assailant using optimization functions (FiRAO-PG) in wireless sensor network. Mathematical Problems
in Engineering, 2015, 2015. doi:10.1155/2015/594345
234
Fruit Fly Optimization-Based Adversarial Modeling for Securing Wireless Sensor Networks (WSN)
Tague, P., Slater, D., Rogers, J., & Poovendran, R. (2008, April). Vulnerability of network traffic under
node capture attacks using circuit theoretic analysis. In IEEE INFOCOM 2008-The 27th Conference on
Computer Communications (pp. 161-165). IEEE 10.1109/INFOCOM.2008.41
Wang, Z., Zhou, C., & Liu, Y. (2017). Efficient hybrid detection of node replication attacks in mobile
sensor networks. Mobile Information Systems.
Zhu, Q., Bushnell, L., & Başar, T. (2012, December). Game-theoretic analysis of node capture and clon-
ing attack with multiple attackers in wireless sensor networks. In 2012 IEEE 51st IEEE Conference on
Decision and Control (CDC) (pp. 3404-3411). IEEE. 10.1109/CDC.2012.6426481
Collusion Resistant: It is the property of key management in which the keys generated by key server
are distinct. No two nodes are assigned same set of cryptographic keys.
Key Management: It is the process of creating, distributing to the senor nodes before they are de-
ployed in sensor field. Later, it also includes the key refreshment process where cryptographic keys are
changed to make the scheme secure against node capture attack.
Key Predistribution: It is one of the phase of the key management in which the cryptographic keys
are created by the key distribution server. The generated keys should be collusion resistant.
Node Capture Attack: It is one of the prominent attack in WSN due to its hostile placement. During
this attack, the adversary physically captures the node and steals the keying information stored in them
by the network designer. It is used to launch future attacks in network.
Resistance Against Node Capture: It is the ability of the network to survive during node capture
attack. It is the number of valid links of non-captured nodes are compromised when n nodes are captured
by the adversary.
235
236
Chapter 13
Cybersecurity Risks Associated
With Brain-Computer
Interface Classifications
Sergio López Bernal
University of Murcia, Spain
ABSTRACT
Brain-computer interfaces (BCIs) have experienced a considerable evolution in the last decade, expanding
from clinical scenarios to sectors such as entertainment or video games. Nevertheless, this populariza-
tion makes them a target for cyberattacks like malware. Current literature lacks comprehensive works
focusing on cybersecurity applied to BCIs and, mainly, publications performing a rigorous analysis of the
risks and weaknesses that these interfaces present. If not studied properly, these potential vulnerabilities
could dramatically impact users’ data, service availability, and, most importantly, users’ safety. Because
of that, this work introduces an evaluation of the risk that each BCI classification already defined in the
literature presents to raise awareness between the readers of this chapter about the potential threat that
BCIs can generate in the next years if comprehensive measures, based on standard mechanisms, are not
adopted. Moreover, it seeks to alert academic and industrial stakeholders about the impact these risks
could have on future BCI hardware and software.
DOI: 10.4018/978-1-7998-7789-9.ch013
Copyright © 2022, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
INTRODUCTION
Application Scenarios
Brain-Computer Interfaces (BCIs) are bidirectional devices that allow communication between the brain
and external systems, such as computers. On the one hand, BCI systems permit the acquisition of neuronal
data, study the status of the brain, and control external devices such as wheelchairs or prosthetic devices
(Lebedev et al., 2017). On the other hand, they are used for neurostimulation procedures to stimulate
targeted regions of the brain (Edwards et al., 2017). These separate approaches can be unified under
the concept of bidirectional BCIs, which can alternate recording and stimulation actions to verify that
the stimulation actions have the desired effect within the brain or offer feedback to BCI users through
stimulation (Rao, 2019).
BCI technologies are mainly used in medical scenarios for the diagnosis and treatment of neurological
diseases. Focusing on the acquisition perspective, BCIs are extensively used to detect various conditions,
such as epilepsy, which allow the detection of anomalous neural activity (Sowndhararajan et al., 2018).
They have also been used as mental speech systems, where the BCI can analyze the user’s neural activ-
ity to determine each spelled letter (Guan et al., 2004). Additionally, technologies such as fMRI based
on magnetic fields allow the visualization of the whole brain to detect damaged tissue, extensively used
nowadays in most hospitals worldwide. Regarding neural stimulation, a wide variety of technologies is
currently accepted by health agencies such as the American FDA to treat different illnesses. For example,
Deep Brain Stimulation (DBS) is a BCI technology used to treat several conditions, such as Parkinson’s
Disease, essential tremor, or obsessive-compulsive disorders (Edwards et al., 2017).
Although BCIs are commonly used for medical purposes, these systems have gained popularity in
other economic sectors. One of the most relevant is the military scenario, where the application of BCIs
is studied in multiple scenarios (Binnendijk et al., 2020). One of these application scenarios is systems
able to monitor soldiers’ mental state, assessing the cognitive and emotional response against complex
situations. They are also being researched to augment soldiers’ mental capabilities in physical, cognitive,
and emotional dimensions. Additionally, they could serve as systems to mentally transmit commands
between individuals, improving the telecommunication systems on the battlefield. Finally, these devices
are interesting for controlling exoskeletons, aiming to restore lost motor functionality or even improve
natural human strength.
Another emerging application sector for BCIs is the entertainment and video game industries, where
BCI systems have been used for controlling the avatar of the game with the mind (Ahn et al., 2014). The
application of BCI technologies could generate a revolution in this field, similar to introducing virtual
reality technologies in recent years. Despite these recreational purposes, the application of BCIs in video
games also positively benefits neurorehabilitation therapies, where these systems can improve lost motor
abilities (McMahon et al., 2018).
Moreover, new technologies and application scenarios have emerged in recent years, presenting
considerable engineering challenges. First, invasive BCIs are moving to the miniaturization of their
electrodes, aiming to reduce the damage caused by surgical procedures, and a better resolution and
coverage of the brain (Musk, 2019). Furthermore, BCIs are incipiently used to allow users to access
the Internet with their minds (Saboor et al., 2018) or even allow direct mental communication between
subjects (Pais-Vieira et al., 2013).
237
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
The evolution and diversification of these technologies have derived in a differentiation of BCIs based
on multiple classifications. One of these classifications is the invasiveness of BCI devices, differentiating
those placed within the skull from those applied externally (Bonaci et al., 2015a; Lebedev et al., 2017).
A second classification is based on the BCI design, which indicates how the acquisition or stimulation
process is initiated, either by the user or the BCI (Wahlstrom et al., 2016; Ramadan et al., 2017). Based
on these differentiations, multiple technologies have emerged in recent years, aiming to satisfy require-
ments not covered by the existing ones for both acquisition and stimulation procedures. The technology
used represents the third classification analyzed in this chapter (Ramadan et al., 2017; Polanía et al.,
2018). Another interesting classification consists in the synchronization of the BCIs, indicating who
controls the processes (An et al., 2016; Ramadan et al., 2017). The last studied classification lies in the
application scenario, such as neuromedical or gaming and entertainment (Li et al., 2015).
Motivation
Although BCI technologies present enormous benefits and future potential, in the past few years, the
academic literature has addressed cybersecurity in BCIs, detecting attacks that can have a negative
effect on these technologies. These works have identified cryptographic attacks aiming to disrupt the
encryption mechanisms used to protect the transmitted data between BCI devices and computers, be-
ing essential to implement robust cryptographic systems (Gupta et al., 2016). They also indicated that
jamming attacks focused on emitting electromagnetic noise could prevent the electrodes from acquiring
neural data (Bonaci et al., 2015a; Ienca et al., 2016). Additionally, Bonaci et al. (2015a) detected that
attackers could use malware strategies to affect data integrity and data confidentiality, compromising
the normal functioning of the system. However, these works only focused on particular aspects of BCI
systems, without an in-depth coverage of the situation. Based on this lack of comprehensive studies,
Lopez-Bernal et al. (2019) reviewed the state of the art of cybersecurity applied to BCIs, analyzing
attacks, impacts, and countermeasures applicable over both the BCI functioning cycle and common
architectural deployments. However, that research does not cover the cybersecurity analysis from the
perspective of BCI classifications. Based on that, this chapter detects an opportunity in the academic
literature to perform a risk evaluation of existing BCI systems.
Contributions
Considering the absence of literature analyzing the cybersecurity aspects related to BCI classifications,
the main contribution of this chapter is a review of the most critical risks associated with each particu-
lar classification from the data integrity, data confidentiality, data and service availability, and safety
perspectives. This classification analysis highlights the most relevant cybersecurity impacts of each BCI
family, aiming to alert about the vulnerabilities present in BCI technologies.
The remainder of this chapter is structured as follows. The Background section presents the current
state of the art regarding cybersecurity on BCIs, and introduces the most common BCI classifications
defined in the literature. After that, the Risk assessment of BCI classifications section presents an analy-
sis of the risks existing for BCI classification. Section Future research directions highlights the current
trend and future evolution of BCIs, indicating possible opportunities in this area. Finally, the Conclusion
section presents a summary of the chapter.
238
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
BACKGROUND
This section first focuses on the analysis of cyberthreats proposed by the current academic literature.
After that, it presents the most common BCI classifications proposed in the literature, indicating the
BCI families included in each classification.
Cybersecurity on BCIs
Research focusing on cybersecurity applied to BCIs has gained importance in recent years, where the
literature has documented threats against these technologies. Among the identified cyberattacks, Mar-
tinovic et al. (2012) documented that the presentation of malicious visual stimuli to BCI users using
the P300 paradigm could extract sensitive information such as PIN numbers, the user’s month of birth,
area of living, or bank-related data. Frank et al. (2017) extended this study, considering the idea of
subliminal visual stimuli, where the user is not aware of the attack, and determining that this approach
could be feasible.
Li et al. (2015) and Landau et al. (2020) identified the applicability of replay and spoofing attacks
over BCI systems, where the attacker acquires legitimate neuronal signals from a user, later retransmitting
them to the medium. These could be particularly damaging in BCI authentication scenarios, where the
attacker obtains the neural signals able to authenticate a user, used later to gain control over the protected
assets. Additionally, the literature has detected the feasibility of jamming attacks (Ienca et al., 2016;
Landau et al., 2020), preventing the BCI from acquiring neural signals from the brain by transmitting
electromagnetic noise to the medium.
One of the most damaging attacks in computer systems is malware (Gupta et al., 2018), which is
also one of the most harmful threats over BCI systems. Bonaci et al. (2015a) documented that current
BCI applications and app stores do not implement security standards, not consider privacy or security
measures. Based on that, a malicious developer could create an application containing malware, defining
two attack strategies. The first one would focus on hijacking the legitimate components and algorithms
of the BCI, changing the behavior of the feature extraction and decoding phases of the BCI functioning
cycle. The second attack could focus on adding or replacing legitimate BCI components. In summary,
malware attacks could alter the integrity of the BCI, making them vulnerable to malicious actions.
Inspired by the previous attacks, López Bernal et al. (2020a) identified vulnerabilities over emerg-
ing BCI solutions focused on stimulating individual neurons, such as Neuralink. They indicated that,
based on these vulnerabilities, an attacker could take advantage of these vulnerabilities to control the
BCI to disrupt the neurostimulation and alter spontaneous neural behavior. They defined the concepts of
Neuronal Flooding and Neuronal Scanning cyberattacks as mechanisms able to disrupt neural behavior,
inspired by computer communication attacks and mechanisms. The experimentation presented indicated
that these threats can reduce the number of neuronal spikes produced, generating an increment in the
temporal dispersion of these spikes as well.
Although the literature has addressed specific cybersecurity threats against BCIs, there was a neces-
sity for academic works gathering all the knowledge about the topic. Based on that, López Bernal et al.
(2019) introduced a fist identification of cyberattacks and countermeasures applicable to the phases of
the BCI cycle. After that, López Bernal et al. (2020b) comprehensively reviewed the state of the art of
cybersecurity on BCIs, analyzing the attacks applicable to each phase of the BCI functioning cycle, the
impacts that they generate, and the possible countermeasures to mitigate them. Additionally, that work
239
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
also reviewed the attacks, impacts, and countermeasures of the most common architectural deployments.
In summary, the authors determined a lack of comprehensive works and solutions addressing cyberse-
curity issues in the BCI field.
BCI Classifications
This subsection presents the different classifications studied in the chapter, previously defined in the
academic literature, studying the design of the BCIs, the technology that they use, their level of invasive-
ness, the synchronization approach used between the user and the BCI, and finally, the BCI application
scenario.
Design of BCIs
The first classification studied in this work is the design of BCIs, which indicates who is responsible
for initiating the neural data acquisition process and how the process is performed. In other words, this
classification indicates if the user or the BCI triggers the process and how the interaction is done to
perform the intended actions (Zander et al., 2010; Wahlstrom et al., 2016; Hong et al., 2017; Lebedev et
al., 2017; Ramadan et al., 2017). According to this categorization, four BCI families have been defined
in the literature: active, passive, reactive, and hybrid.
In active BCIs, also known as independent (Lebedev et al., 2017) or spontaneous (Ramadan et al.,
2017), users intentionally initiate a predefined action that generates brain activity, such as imaging limb
movements, which is then captured by the BCI (Zander et al., 2010; Lebedev et al., 2017). An example
of an action is the imagination of limb movements (Lebedev et al., 2017). Active BCIs have been used
by Gilja et al. (2012) for the control of a computer cursor in rhesus monkeys.
Passive BCIs, in contrast to active BCIs, focus on the acquisition of spontaneous and non-evoked brain
activity, typically generated during complex real-world tasks, such as the actions performed by a pilot
during a flight (Zander et al., 2010, Wahlstrom et al., 2016). Passive BCIs have been used to measure
mental states such as attention, stress, workload, or emotions (Aricò et al., 2018).
The third family is termed Reactive BCIs, also identified as dependent (Lebedev et al., 2017) or
evoked (Ramadan et al., 2017). This family depends on external stimuli presented to the users and the
neural responses generated by their brains as a response. Reactive BCIs are used, for example, to detect
situations in which the user recognizes known external stimuli from a set of unknown stimuli.
Finally, Hybrid BCIs can be considered from two perspectives. On the one hand, they are BCI systems
that receive different types of brain signals as input (Lebedev et al., 2017). Focusing on this approach,
Ramadan et al. (2017) and Hong et al. (2017) showed different possibilities and their purpose. For
example, a combination of EEG and EMG to improve accuracy and enhance application performance.
On the other hand, Wahlstrom et al. (2016) defined hybrid BCIs as a combination of at least one ac-
tive, passive, or reactive BCI system with non-BCI technologies to improve system performance (e.g.,
a combination of an active BCI and a finite-state automaton to control a robot).
BCI Technology
The technology used by the BCI is the second classification, where two additional subclassifications
arise depending on if they are intended for acquiring neural data or for brain stimulation. The most rep-
240
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
resentative technologies considering the acquisition of brain waves are Electroencephalography (EEG),
Functional Magnetic Resonance Imaging (fMRI), Magnetoencephalography (MEG), Electrocorticog-
raphy (ECoG), and neural dust. On the other hand, focusing on brain stimulation techniques, the most
relevant ones are Transcranial Magnetic Stimulation (TMS), Transcranial Electrical Stimulation(tES),
Transcranial Focused Ultrasound (tFUS), Deep Brain Stimulation (DBS), and neural dust (Tyler et al.,
2017; Polanía et al., 2018).
Related to neural activity recording, EEG is a non-invasive technology that uses electrodes placed
on the scalp (Ramadan et al., 2017). This technology highlights for being easy to use, cheap, portable,
and with a high temporal resolution, in the order of milliseconds (vanGerven et al., 2009; Ramadan et
al., 2017). It is the most widely used non-invasive technology, mainly in video games and entertainment
(Ahn et al., 2014; McMahon et al., 2018). However, it has a poor spatial resolution, and the acquired
brain waves can be distorted by other neural activity, such as the movements of muscles and eyes, known
as artifacts (vanGerven et al., 2009).
fMRI is another non-invasive technology used to measure the variation of blood hemoglobin con-
centrations during brain activity. It has better spatial resolution than EEG and MEG, identifying active
zones throughout the brain. However, it has a low temporal resolution, between one and two seconds
(vanGerven et al., 2009; Lebedev et al., 2017; Ramadan et al., 2017). The resolution of fMRI is also
affected by head movements (Ramadan et al., 2017). This technology is useful in clinical scenarios to
treat neurological conditions such as stroke or mental disorders, although it does not apply to everyday
use (Lebedev et al., 2017).
Continuing with non-invasive recording technologies, MEG uses functional neuroimaging over mag-
netic fields produced by the electric current generated by cortical neurons. This method has an excellent
temporal and spatial resolution. Despite these benefits, MEG can only be used in magnetically shielded
installations, and it is not portable. This technology is widely used to detect regions with abnormal brain
functions and in tetraplegic and stroke patients (Lebedev et al., 2017; Ramadan et al., 2017).
Finally, ECoG is a partial-invasive method in which a grid of electrodes is placed on the brain sur-
face (Lebedev et al., 2017). It offers an SNR superior to non-invasive systems, such as EEG, and higher
spatial and temporal resolution. Besides, it allows better detection of high-frequency oscillatory activity
(vanGerven et al., 2009). Despite the above advantages, ECoG cannot be used to detect single-neuron
spikes (Lebedev et al., 2017), and it is very challenging to use these devices outside an operating room
(Ramadan et al., 2017).
Regarding the stimulation of neurons, TMS is a technology that generates electrical fields within the
brain, reaching the cortex and aiming to modulate brain activity and behavior. This technology obtained
FDA approval in 2018 to treat depression and headaches (Tyler et al., 2017). TMS has also been used
for testing dynamic communication between interconnected areas of the brain (Polanía et al., 2018) and
cognitive aging (Gomes-Osman et al., 2018). Although TMS has a good temporal resolution, it presents
an inadequate spatial precision (Lebedev et al., 2017).
tES is another stimulation technique that uses weak, painless currents applied to the scalp (Polanía et
al., 2018). It can be based on direct current stimulation (tDCS) or alternating current stimulation (tACS).
tDCS is a simple method that stimulates the cortex and affects relatively large areas, presenting a low
spatial and temporal resolution. In contrast, tACS presents good temporal precision (Polanía et al., 2018).
It has been reported that tES can enhance and perturb cognitive processes, such as creative problem
solving or working memory, when applied to different brain regions. Furthermore, it can improve work-
241
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
ing memory performance and motor behavior (Tyler et al., 2017; Polanía et al., 2018). Although these
technologies are promising, they are not mature enough for their use in humans in terms of reliability
and reproducibility (Tyler et al., 2017; Bikson et al., 2018).
tFUS is a novel neuromodulation technique that offers a high spatial resolution, being the only non-
invasive technology able to penetrate the skull and stimulate specific circuits deep in the brain (Kubanek
et al., 2018, Legon et al., 2018). It has been used to stimulate the activity of cortical, thalamic, and hip-
pocampal circuits in animals (Tyler et al., 2017), and it may be useful to identify and treat neurological
and psychiatric disorders in humans, such as neuropathic pain or depression, due to its potential to induce
plastic changes in aberrant brain circuits (Kubanek et al., 2018).
Finally, DBS is an invasive neurostimulation technique that involves a surgical procedure for im-
planting electrodes deep within the brain. This invasiveness provides DBS with acceptable spatial and
temporal resolution. Focusing on their functioning, the implanted device sends electric currents into
targeted subcortical areas to increase, suppress, or distort neural activity. This method has been used to
treat conditions such as Parkinson’s disease, dystonia, and chronic pain syndromes (Denning et al., 2009;
Ienca et al., 2016; Khabarova et al.; 2018, Kubanek et al., 2018). Despite the benefits of DBS, the associ-
ated surgeries required may have complications such as infection or hemorrhage (Kubanek et al., 2018).
Neural nanonetworks comprise several technologies to record and stimulate neural activity through
the use of nanodevices. In particular, neural dust is a solution to acquire neural data relying on nanode-
vices allocated into the cortex, beneath the skull, and below the dura mater (Seo et al., 2013; Lebedev
et al., 2017). An interrogator is powered by an external transceiver using radio frequency power transfer
(Seo et al., 2013), and it establishes wireless power and communication with the neural dust using ultra-
sounds (Seo et al., 2013, Neely et al., 2018). This nanotechnology offers some advantages, as it does not
use microelectrode shafts that can damage the nervous tissue, it records very enclosed areas within the
brain, and it can work as a closed-loop system based on real-time adaptation, offering high spatial and
temporal resolution (Lebedev et al., 2017). It is also an interesting alternative to electromagnetic systems
due to its lower attenuation and higher efficiency (Neely et al., 2018). Wirdatmadja et al. (2017) used
the neural dust model defined by Seo et al. (2013) to propose a stimulation system based on optogenetic
nanonetworks and the definition of different firing patterns (e.g., brain regions, frequencies, temporal
synchronization) to interact with the devices. Zheng et al. (2019) developed an implantable device based
on optogenetic stimulation for peripheral nerves, focused on activating limb muscles. Lee et al. (2019)
defined the concept of Neurograin sensors, implementing a network of nanodevices that uses wireless
energy harvesting, and validated in both ex vivo and in vivo rodent tests. Despite the advantages of
these technologies, they are experimental, and they have not yet been tested in humans (Lebedev et al.,
2017). Additionally, the architecture of neural dust has been identified as vulnerable by López Bernal
et al. (2020a).
Level of Invasiveness
The invasiveness level classification indicates whether the BCI device is implanted in the user’s body or
placed externally. This classification has been widely studied in the literature (vanGerven et al., 2009;
Bonaci et al., 2015a; Li et al., 2015; Vaid et al., 2015; Ienca et al., 2016; Wahlstrom et al., 2016; Frank
et al., 2017; Ramadan et al., 2017; Lebedev et al., 2017; Polania et al., 2018), where the following three
families have been proposed: invasive, partial-invasive and non-invasive.
242
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Invasive systems require a neurosurgery process involving opening the scalp, the skull and placing
the BCI components in the brain tissue to record or stimulate neurons. This technology has been used
mainly in the medical field because it allows measuring neural activity with very little noise. In the last
decades, they allow direct interaction with the brain, enabling the stimulation of individual neurons
(Lebedev et al., 2017; Wirdatmadja et al., 2017).
The second family of this category is called partial-invasive, and the BCI components are placed on
the brain surface without penetrating the nervous tissue. This family is used in medicine, for example,
with subdural electrodes aiming to identify the location of epileptic seizures (vanGerven et al., 2009;
Yang et al., 2014). This type of BCIs has lower temporal and spatial resolution than the previous one,
affecting its applicability in some application scenarios.
Finally, non-invasive BCIs are applied outside the skull, directly on the scalp. They present lower
temporal and spatial resolution than the previous two families due to the attenuation and filter provoked
by the bone and skin. However, they have an essential role in the health field, where non-invasive neural
stimulation systems are gaining popularity (Ramadan et al., 2017). Additionally, these technologies are
the most extended systems due to their simplicity and applicability in entertainment scenarios, where
final users benefit from their advantages (Li et al., 2015; Ramadan et al., 2017).
Synchronization
The fourth classification considered in this work is the synchronization of BCIs, focused on the interac-
tion between the BCIs and the users. It determines who controls the recording and stimulation processes
and in which time slots. Based on that, two families of BCIs have been documented in the literature (An
et al., 2016; Ramadan et al., 2017): synchronous and asynchronous.
In synchronous (or cue-paced) BCI systems, the interaction between the user and the BCI occurs
during specific periods (Ramadan et al., 2017). This planning is imposed by the BCI, which controls the
communication. Outside these periods, the BCI is not able to communicate with the user. They are easier
to implement than asynchronous BCIs, but they are not suitable for acquiring users’ mental intentions
(An et al., 2016). Bentabet et al. (2016) used synchronous BCIs to control domotic devices, extracting
features from P300 waves.
In contrast, in asynchronous (or self-paced) systems, users can generate brain signals at any time, and
the BCI will react to these events. Ramadan et al. (2017) highlighted the complexity of detecting idle
states and proposed using a button to activate or deactivate the acquisition of stimuli. An et al. (An et
al., 2016) proposed an asynchronous BCI to control a virtual avatar in a game. In this game, the avatar
competes with other users in a race, running continuously and, when a control command is received, an
action on the avatar is performed. As can be seen, these systems depend on the moment and the action
performed by the user, without control imposed from the BCI.
Usage Scenario
The application scenario is another well-known BCI classification. It was proposed by Li et al. (2015),
and they highlighted the following four types: neuromedical, user authentication, gaming and entertain-
ment, and smartphone-based BCIs.
The field of neuromedical applications has been the center of research in BCI for decades. The appli-
cations developed within this field range from the control of prosthetic limbs and wheelchairs (Denning
243
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
et al., 2009; Li et al., 2015; Ienca et al., 2016) to the use in brain stimulation procedures (Lebedev et
al., 2017; Wirdatmadja et al., 2017; Polanía et al., 2018). Besides, Chaudhary et al. (2017) used BCIs to
establish a simple communication system with wholly paralyzed patients. Nowadays, current research
on BCIs focuses on novel mechanisms and technologies to analyze and stimulate the brain.
The second scenario is the use of BCIs as authentication systems. The authentication process consists
in recording the user’s brain waves while performing a previously established task. The acquired neural
data is then validated against the entity, containing the authentication data to validate the user. Brain sig-
nals are an excellent biometric element since each brain generates unique patterns (Bonaci et al., 2015a,
Takabi et al., 2016b), and the mental action that triggers the authentication process can be modified,
enabling an adaptive and flexible authentication mechanism. Finally, brain signals can be easily affected,
and it is difficult to reproduce them under the effects of stress, anxiety, or drugs (Martinovic et al., 2012).
The third family, gaming and entertainment, arose due to the utility of BCIs in the video game indus-
try to control the actions transmitted to the game with the mind. The development tasks allowing their
creation have been facilitated by using standard APIs (Takabi et al., 2016b). Ahn et al. (2014) reviewed
BCI games, highlighting games such as Bacteria Hunt, and performed a survey involving researchers,
game developers, and users. McMahon et al. (2018) focused on virtual reality and created a low-cost
open-source development environment prototype for BCI games.
Finally, the smartphone-based BCIs are based on the relationship between BCIs and user applications
stored in smartphones, where it is the most common usage scenario in commercial BCI brands, such as
Emotive or Neurosky (Takabi et al., 2016b). Additionally, Pycroft et al. (2016) identified that computa-
tional devices, such as smartphones, introduce new risks and security problems. In this sense, Li et al.
(2015) indicated that most of the issues existing in smartphones could also be applied to smartphone-
based BCI applications. For example, gain access to data stored in the SD card, or transfer sensitive data
to remote servers using malware attacks.
This section presents a quantitative analysis of the risks present in each classification introduced in the
previous section, from the data integrity, data confidentiality, data and service availability, and safety
perspectives. To facilitate the lecture of this section, Figure 1 presents a summary of the classifications
studied, also included the different families included in each classification.
Design of BCIs
Regarding the cybersecurity risks associated with each family of this classification, this chapter identifies
malicious external stimuli as the most damaging ones. They are exploited by misleading stimuli attacks
(López Bernal et al., 2020b) by presenting malicious inputs to the user or the BCI to obtain a benefit.
An example of this benefit is the use of subliminal visual stimuli to generate specific brain signals that
imply sensitive data leakages, such as acquiring thoughts or personals beliefs. In this context, Wahl-
strom et al. (2016) indicated that if users with active BCIs can stop the BCI functioning and they have
given consent to acquire such data, there is no risk of confidentiality attacks. However, the authors of
this chapter identify that misleading stimuli attacks applied to these BCIs generate data confidentiality
concerns, as the conditions mentioned above do not prevent BCIs from suffering vulnerabilities.
244
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Additionally, Wahlstrom et al. (2016) documented that passive BCIs are at risk of data confidentiality
attacks since users do not control the BCI. Moreover, they highlighted that in reactive BCIs, confiden-
tiality issues are unlikely to arise if the neural activity is filtered, users are in a controlled environment,
and consent to the procedure. In contrast, based on the subliminal visual principles mentioned above,
misleading stimuli attacks have been detected in the literature for this kind of BCIs (Martinovic et al.,
2012; Frank et al., 2017). Related to hybrid BCIs, Wahlstrom et al. (2016) identified that the risk of these
technologies is the combination of the risks of each of their parts. Considering these aspects, the authors
of this chapter also detect data integrity and data availability concerns in all BCI designs since mislead-
ing stimuli attacks can alter the data acquired by the BCI or even disrupt the data acquisition process.
Besides, the authors identify safety issues generated from these problems. Based on the above concerns,
hybrid BCIs have the highest risk, followed by reactive BCIs, passive BCIs, and finally active BCIs.
Figure 2 summarizes the risks described in this classification. Each family has been represented by
a line color with a particular line style, while a number indicates the risks documented in this section.
The severity of these risks has been considered based on four different types of impacts (integrity,
confidentiality, availability, and safety), represented in the vertex of each radar chart. This severity is
quantified on a scale between zero and three, where a zero value defines an absence of risk, and three
represents a critical high risk.
245
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Figure 2. Cybersecurity and safety risks associated with the design of BCIs
BCI Technology
The authors identify that both temporal and spatial resolutions generate significant cybersecurity risks.
In particular, BCI technologies such as DBS, neural dust, ECoG, or MEG present higher risks in data
confidentiality and safety than those with lower resolutions, like EEG, fMRI, or TMS. It is also impor-
tant to highlight that the invasiveness of these methods also impacts the severity of the risks associated
with each family. Because of that, DBS and neural dust have a higher risk than ECoG due to their inva-
siveness, and all of them are more damaging than MEG, as it is a non-invasive technology. In contrast,
technologies with low resolutions present concerns on data and service availability since they transmit
a reduced amount of data affected more easily by electromagnetic interference.
According to the inherent functioning of acquisition and stimulation systems, two more risks arise.
On the one hand, since acquisition technologies aim to record neural data, they generate data integrity
and confidentiality risks, where attacks can impersonate or gather users’ neural data. On the other hand,
stimulation systems mainly present safety issues, where attackers can cause brain damage. Several works
in the literature review the safety concerns of stimulation technologies. According to Glannon et al.
(2014), ECoG has a risk of infection and hemorrhage, and the microelectrode arrays used present poten-
tial problems of biocompatibility with neural tissue. Besides, Polanía et al. (2018) indicated that TMS
pulses applied to particular areas could induce suppression of visual perception or speech arrest, which
serve as an opportunity for attackers. Finally, the authors of the chapter identify electromagnetic noise
as a risk directed over the physical aspects of non-invasive transmission systems. Technologies such as
EEG acquire electrical currents, while fMRI and MEG acquire magnetic fields emitted from the brain.
These specific aspects can serve as an opportunity for attackers to override the legitimate information
generated by the brain, creating concerns in data integrity and availability.
246
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Finally, Figure 3 indicates the risks that affect each of the BCI technologies documented in this clas-
sification. It is important to note that the invasiveness of these technologies also influences the risks
associated with precision and electromagnetic noise. Because of that, invasive systems offer higher risks
than non-invasive methods.
Figure 3. Cybersecurity and safety risks associated with the BCI technology
Level of Invasiveness
Analyzing the previous three families from the cybersecurity point of view, the authors have identified
that two of the most severe risks are temporal and spatial resolutions. In this context, BCIs with higher
spatial resolution can access more precise neural data or stimulate more specific brain regions than those
with lower precision (Lebedev et al., 2017; Ramadan et al., 2017). A high temporal resolution allows
attackers to perform more complex attacks since the communication delay is reduced. In this context,
invasive systems have access to neural-level data, whereas less invasive systems acquire aggregated data
with less resolution. That increase of precision generates concerns in data confidentiality, where systems
247
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
with higher resolution have access to more precise and detailed information (e.g., thoughts or beliefs).
Besides, the BCI precision can impact users’ physical safety, where a high precision can increase the
damage of attacks during neurostimulation processes. However, a reduction of the precision produces
the transmission of a reduced quantity of data that can be insufficient for the correct functioning of BCIs
in specific scenarios, impacting service availability. The level of intrusiveness is another risk detected
in the literature. Invasive and partial-invasive BCIs are at risk of tissue damage, infection, and rejection
due to the surgical procedure required to place them (Li et al., 2015; Waldert et al., 2016; Lebedev et
al., 2017; Ramadan et al., 2017; Campbell et al., 2018; Pycroft et al., 2018). Also, both present a risk
of degradation in the acquisition and stimulation technologies used, such as electrodes, when used for
long periods (Campbell et al., 2018).
Invasive systems have the highest risk within the temporal and spatial resolutions, followed by par-
tial- and non-invasive BCIs. Invasive systems introduce more severe concerns in terms of users’ physical
integrity than partial-invasive, while non-invasive BCIs are immune to them. Nevertheless, non-invasive
stimulation systems significantly reduce, but not suppress, safety risks (Polanía et al., 2018). At this point,
it is essential to highlight that the majority of attacks on BCIs have been conducted over non-invasive
systems. However, it is motivated by their designs and implementations and not due to problems inher-
ent to the level of invasiveness. Finally, Figure 4 represents the risks detected for this BCI classification.
Figure 4. Cybersecurity and safety risks associated with the invasiveness of BCIs
Synchronization
Considering the cybersecurity risks generated by these two families of BCIs, no risks have been docu-
mented in the literature. However, the authors identify that their main issue is losing control over the
communication between a BCI and its user. In this context, the authors detected that synchronous BCIs,
248
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
which control the communication, originate data integrity and confidentiality concerns, where attackers
taking control over the BCI can gather and alter the neural data. Moreover, Ienca et al. (2018) highlighted
the third parties with access to brain waves could extract private information from subjects. In this sense,
we consider that, although this issue is extensible to all BCI families, synchronous systems could have
a higher impact since more sensitive information could be acquired. Additionally, availability issues
are possible, where attackers disable the data acquisition process, even without the users’ knowledge.
Finally, this lack of control can generate safety impacts, where attackers managing the functionality of
BCIs can produce critical physical harm, such as malicious movements of a wheelchair or damaging
stimulation patterns.
On the other hand, since asynchronous BCIs have fewer decision capabilities and awareness over the
acquisition process, this chapter identifies that this BCI family has a risk of malicious external stimuli
aiming to perform misleading stimuli attacks. These stimuli originate data integrity concerns, where at-
tackers impersonate the neural data generated that is communicated to the BCI. Moreover, they generate
data and service availability issues since these stimuli can impact the acquisition process and, thus, the
normal functioning of the communication. Finally, the authors consider that asynchronous BCIs have a
high risk of confidentiality problems, where the BCIs are constantly acquiring brain activity, and there-
fore sensitive information is gathered. The previous attacks and concerns also generate safety problems.
Considering the above, asynchronous BCIs present greater data integrity and availability issues due
to their more extensive temporal exposition to neural data. However, considering the user’s awareness
and control capabilities over the communication, synchronous BCIs have more serious concerns over
both issues. Although both synchronous and asynchronous BCIs share common concerns, the authors
detect that the first family has a higher risk since attacks over BCIs are more probable than those based
on user neural data impersonation. This situation is summarized in Figure 5, which highlights the dif-
ference of severity between these BCI families.
Figure 5. Cybersecurity and safety risks associated with the synchronization of BCIs
249
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Usage Scenario
The cybersecurity risks of each BCI family vary considerably according to the usage scenario. These
risks generate data confidentiality concerns in medical scenarios, where attackers can gather sensitive
information (Li et al., 2015; Sempreboni et al., 2018). Moreover, users are at risk of physical harm since
BCIs are used to improve patients’ health conditions. Denning et al. (2009) identified safety concerns
based on malicious neural stimulation actions, whereas Li et al. (2015) detected service integrity issues
if users modify the parameters that control prosthetic limbs to gain a personal benefit. The authors of this
chapter also identify that neuromedical scenarios risk the management of sensitive information of patients,
such as their personal data, medical history, and neural activity data, affecting users’ confidentiality.
Li et al. (2015) identified data confidentiality concerns based on the acquisition of the authentication
data regarding authentication scenarios. The authors of the present chapter detect that they are at risk of
malicious external stimuli aiming to alter the neural data used for the authentication process and thus
impact the data integrity and availability. On the other hand, this chapter identifies that the gaming and
entertainment scenario has a risk of malicious external stimuli, as this family is based on audiovisual
systems that serve as an opportunity for attackers to perform adversarial attacks. The authors highlight
data confidentiality issues, where attackers present malicious stimuli to acquire sensitive data taking
advantage of these multimedia resources. The authors also identify that this situation affects data integ-
rity and availability. Finally, smartphone-based scenarios present several risks. First, they rely on sys-
tems with potential problems, such as a lack of updates of the Operating System (OS) and applications
(Martinovic et al., 2012; Bonaci et al., 2015b; Li et al., 2015; Sundararajan, 2017). The authors of the
chapter also detect that the heterogeneity of the hardware, OS, applications used, and versions of each
specific smartphone can also produce cybersecurity risks (StatCounter Global Stats, 2020). Based on
that, Takabi et al. (Takabi et al., 2016b) analyzed several smartphone applications developed with the
NeuroSky platform (NeuroSky, 2020). They detected that some third-party applications required access
to the phone book and permissions to read the call logs, which was not the objective of the applica-
tions, generating confidentiality concerns. Moreover, this chapter detects that this lack of control over
the smartphone elements generates concerns about data integrity, data availability, and safety, where
attackers could perform well-known cyberattacks over the users and their data (Agrawal et al., 2018).
In conclusion, Figure 6 provides an overview of the previous concerns. Due to their inherent critical
actions, the highest risks of neuromedical scenarios are on integrity, confidentiality, and safety issues.
For authentication systems and gaming and entertainment scenarios, the authors consider that integrity,
confidentiality, and availability concerns are equally probable. Finally, smartphone-based scenarios
present all four concerns.
Finally, Table 1 groups all the information described in the section. It indicates, for each classification
and BCI family, the general references that treat relevant concepts associated with each BCI family. The
four concerns analyzed throughout the section are exposed (i.e., integrity, confidentiality, availability,
and safety), where one or more references indicate that the BCI family presents a concern previously
documented in the literature. Moreover, our contribution is indicated with a red X icon (✗). Finally, a
green check icon (✓) indicates that there are no concerns identified.
250
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Figure 6. Cybersecurity and safety risks associated with the BCI usage scenario
The BCI field has presented a considerable development in recent years, moving from medical and re-
search scenarios to commercial ones. This growth defines emerging and promising application scenarios,
such as the connection of BCIs to the Internet, allowing the transfer and storage of neural data to cloud
environments. Another interesting scenario is the communication between brains, known as brain-to-
brain (BtB) communication, where two or more BCI users mentally share information.
Although these advances in the BCI field are promising, they introduce new challenges and risks,
especially by applying well-known cyberattacks. An example of this situation is malware, where attack-
ers could affect the integrity, confidentiality, and availability of data and services, also causing physical
damage to these users. The application of new communication mechanisms and the use of cloud comput-
ing strategies augment the number of elements potentially vulnerable in the architecture of the services,
easing the tasks of finding vulnerabilities and exploiting them to perform malware attacks.
251
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Table 1. Cybersecurity risks and concerns based on BCI classifications, differentiating between those
families that have a lack of cybersecurity issues (✓) and those with identified problems, either documented
in the literature (list of references) or detected by the authors of this chapter (✗).
Cybersecurity concerns
BCI
BCI family Family-related literature Integrity Confidentiality Availability Safety
classification
Zander et al. (2010); Gilja et
al. (2012); Wahlstrom et al.
Active ✗ ✗ ✗ ✗
(2016); Lebedev et al. (2017);
Ramadan et al. (2017)
Zander et al. (2010);
Wahlstrom et al. (2016); Wahlstrom et al.
Design of BCIs
Passive ✗ ✗ ✗
Lebedev et al. (2017); Aricò et (2016)
al. (2018)
Martinovic et al. (2012);
Martinovic et al.
Wahlstrom et al. (2016);
Reactive ✗ (2012); Frank et ✗ ✗
Lebedev et al. (2017);
al. (2017)
Ramadan et al. (2017)
Hong et al. (2017); Wahlstrom Wahlstrom
Wahlstrom et al. Wahlstrom et Wahlstrom
Hybrid et al. (2016); Lebedev et al. et al.
(2016) al. (2016) et al. (2016)
(2017); Ramadan et al. (2017) (2016)
Ramadan et al. (2017);
vanGerven et al. (2009); Ahn
EEG ✗ ✗ ✗ ✓
et al. (2014); McMahon et al.,
(2018)
vanGerven et al. (2009);
fMRI Lebedev et al. (2017); ✗ ✗ ✗ ✓
Ramadan et al. (2017)
Lebedev et al. (2017);
MEG ✗ ✗ ✗ ✗
Ramadan et al. (2017)
vanGerven et al. (2009);
Glannon et
ECoG Lebedev et al. (2017); ✗ ✗ ✓
al. (2014)
Ramadan et al. (2017)
BCI Technology
252
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Table 1. Continued
Cybersecurity concerns
BCI
BCI family Family-related literature Integrity Confidentiality Availability Safety
classification
Lebedev et al. (2017);
Campbell et
Level of invasiveness
This chapter identifies the importance of performing comprehensive risk assessment actions over BCI,
applying well-known methodologies from cybersecurity scenarios. In particular, the authors highlight the
relevance of using standard methodologies such as the one defined by the ISO 27001 to unify the study
of BCI risks and vulnerabilities, allowing to contrast the results obtained by future works in the literature.
This chapter also detects an opportunity to analyze the risks and vulnerabilities of commercial BCI
brands and their products, aiming to extend the risk analysis performed in this chapter. As indicated in
the present work, the literature has identified vulnerabilities in consumer-grade BCI products, although
a comprehensive analysis of the most used BCI devices is required. It is essential to ensure that a con-
siderable amount of BCI technologies in the market are previously reviewed by external cybersecurity
experts, making it difficult to exploit vulnerabilities by potential attackers.
Among the risks detected in this chapter, one of the most common to exploit is the acquisition of
sensitive neural data and acquiring data concerning users’ health status or private data from their medical
records. These devices must be analyzed to verify that they comply with most data protection regulations,
such as the legislation of the United States of America or the General Data Protection Regulation (GDPR)
in the European Union (Huertas et al., 2019; Huertas et al., 2020). If BCI devices are not manufactured
253
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
considering security-by-design and privacy-by-design principles, the possibility of attacks over data
confidentiality alarmingly increases.
CONCLUSION
This chapter has presented the analysis of risk existing in the most common BCI classifications from
the academic literature. The authors have studied these risks from the perspective of data and service
integrity, confidentiality and availability, and BCI users’ physical safety. These risks are motivated by
the inherent vulnerabilities present in each classification and the possibility of applying well-known
cyberattacks, such as malware. Particularly, malware has been identified in the literature as one of the
most damaging threats against BCI, where attackers can control the BCI to perform malicious actions
or serve as a mechanism to obtain sensitive information.
This chapter also highlights that all families within each classification present vulnerabilities, iden-
tifying that the BCI design determined for a device considerably determines the type and severity of the
risks. Particularly, both reactive and hybrid BCIs present higher risks than active and passive approaches.
There are considerable differences in BCI technology as each technology is used in particular scenarios
and solves different problems. In general, this chapter detects that those technologies with higher resolu-
tions for both acquisition and stimulation procedures are the most vulnerable.
The invasiveness of the BCI, also related to the BCI technology classification, has a considerable
influence over the risks of its families, where invasive technologies have the highest impact. The syn-
chronization mechanisms of the BCI, which indicate if the actions are either initiated by the BCI or the
user, are vulnerable, where synchronous BCIs present higher risks in terms of loss of communication
control. Finally, the application scenario present high differences between its families, since each scenario
have specific characteristics and requirements.
ACKNOWLEDGMENT
This work has been partially supported by (a) Bit & Brain Technologies S.L. under the project Cyber-
Brain, associated with the University of Murcia (Spain), by (b) the Swiss Federal Office for Defense
Procurement (armasuisse) with the CyberSpec (CYD-C-2020003) project, and by (c) the University of
Zürich UZH.
REFERENCES
Agrawal, D. P., & Wang, H. (2018). Computer and Cyber Security: Principles, Algorithm, Applications,
and Perspectives (B. B. Gupta, Ed.; 1st ed.). CRC Press. doi:10.1201/9780429424878
Ahn, M., Lee, M., Choi, J., & Jun, S. C. (2014). A Review of Brain-Computer Interface Games and
an Opinion Survey from Researchers, Developers and Users. Sensors (Basel), 14(8), 14601–14633.
doi:10.3390140814601 PMID:25116904
254
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
An, H., Kim, J., & Lee, S. (2016). Design of an asynchronous brain-computer interface for control of a
virtual Avatar. 2016 4th International Winter Conference on Brain-Computer Interface (BCI) (pp. 1-2).
10.1109/IWW-BCI.2016.7457463
Aricò, P., Borghini, G., Di Flumeri, G., Sciaraffa, N., & Babiloni, F. (2018). Passive BCI beyond the
lab: Current trends and future directions. Physiological Measurement, 39(8), 08TR02. Advance online
publication. doi:10.1088/1361-6579/aad57e PMID:30039806
Bentabet, N., & Berrached, N. (2016). Synchronous P300 based BCI to control home appliances. 2016
8th International Conference on Modelling, Identification and Control (ICMIC) (pp. 835-838). 10.1109/
ICMIC.2016.7804230
Bikson, M., Brunoni, A. R., Charvet, L. E., Clark, V. P., Cohen, L. G., Deng, Z. D., Dmochowski, J.,
Edwards, D. J., Frohlich, F., Kappenman, E. S., Lim, K. O., Loo, C., Mantovani, A., McMullen, D. P.,
Parra, L. C., Pearson, M., Richardson, J. D., Rumsey, J. M., Sehatpour, P., ... Losanby, S. H. (2018). Rigor
and reproducibility in research with transcranial electrical stimulation: An NIMH-sponsored workshop.
Brain Stimulation, 11(3), 465–480. doi:10.1016/j.brs.2017.12.008 PMID:29398575
Binnendijk, A., Marler, T., & Bartels, E. M. (2020). Brain-Computer Interfaces: U.S. Military Applica-
tions and Implications, An Initial Assessment. RAND Corporation. doi:10.7249/RR2996
Bonaci, T., Calo, R., & Chizeck, H. J. (2015). App Stores for the Brain: Privacy and Security in Brain-Com-
puter Interfaces. IEEE Technology and Society Magazine, 34(2), 32–39. doi:10.1109/MTS.2015.2425551
Bonaci, T., Herron, J., Matlack, C., & Chizeck, H. J. (2015). Securing the Exocortex: A Twenty-First
Century Cybernetics Challenge. IEEE Technology and Society Magazine, 34(3), 44–51. doi:10.1109/
MTS.2015.2461152
Campbell, A., & Wu, C. (2018). Chronically Implanted Intracranial Electrodes: Tissue Reaction and
Electrical Changes. Micromachines, 9(9), 430. doi:10.3390/mi9090430 PMID:30424363
Chaudhary, U., Xia, B., Silvoni, S., Cohen, L. G., & Birbaumer, N. (2017). Brain–Computer Interface–
Based Communication in the Completely Locked-In State. PLoS Biology, 15(1), 1–25. doi:10.1371/
journal.pbio.1002593 PMID:28141803
Chizeck, H. J., & Bonaci, T. (2014). Brain-Computer Interface Anonymizer (Patent application
US20140228701A1). United States Patent and Trademark Office.
Denning, T., Matsuoka, Y., & Kohno, T. (2009). Neurosecurity: Security and privacy for neural devices.
Neurosurgical Focus FOC, 27(1), E7. doi:10.3171/2009.4.FOCUS0985 PMID:19569895
Edwards, C., Kouzani, A., Lee, K. H., & Ross, E. K. (n.d.). Neurostimulation Devices for the Treatment
of Neurologic Disorders. Mayo Clinic Proceedings, 92(9), 1427-1444. doi:10.1016/j.mayocp.2017.05.005
Frank, M., Hwu, T., Jain, S., Knight, R. T., Martinovic, I., Mittal, P., Perito, D., Sluganovic, I., & Song, D.
(2017) Using EEG-Based BCI Devices to Subliminally Probe for Private Information. Proceedings of the
2017 on Workshop on Privacy in the Electronic Society - WPES ’17, 133-136. 10.1145/3139550.3139559
255
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Gilja, V., Nuyujukian, P., Chestek, C., Cunningham, J. P., Yu, B. M., Fan, J. M., Churchland, M. M.,
Kaufman, M. T., Kao, J. C., Ryu, S. I., & Shenoy, K. V. (2012). A high-performance neural prosthesis
enabled by control algorithm design. Nature Neuroscience, 15(12), 1752–1757. doi:10.1038/nn.3265
PMID:23160043
Glannon, G. (2014). Ethical issues with brain-computer interfaces. Frontiers in Systems Neuroscience,
8, 136. doi:10.3389/fnsys.2014.00136 PMID:25126061
Gomes-Osman, J., Indahlastari, A., Fried, P. J., Cabral, D. L. F., Rice, J., Nissim, N. R., Aksu, S.,
McLaren, M. E., & Woods, A. J. (2018). Non-invasive Brain Stimulation: Probing Intracortical Circuits
and Improving Cognition in the Aging Brain. Frontiers in Aging Neuroscience, 10, 177. doi:10.3389/
fnagi.2018.00177 PMID:29950986
Guan, C., Thulasidas, M., & Wu, J. (2004). High performance P300 speller for brain-computer interface.
IEEE International Workshop on Biomedical Circuits and Systems (pp. S3/5/INV-S3/13). 10.1109/
BIOCAS.2004.1454155
Gupta, B., Agrawal, D. P., & Yamaguchi, S. (2016). Handbook of Research on Modern Cryptographic
Solutions for Computer and Cyber Security. IGI Global. doi:10.4018/978-1-5225-0105-3
Gupta, D., Gupta, B., & Mishra, A. (2018). Identity Theft, Malware, and Social Engineering in Dealing
with Cybercrime. In B. B. G. Gupta (Ed.), Computer and Cyber Security. CRC Press.
Hong, K., & Khan, M. J. (2017). Hybrid Brain–Computer Interface Techniques for Improved Classifi-
cation Accuracy and Increased Number of Commands: A Review. Frontiers in Neurorobotics, 11, 35.
doi:10.3389/fnbot.2017.00035 PMID:28790910
Huertas Celdrán, A., Gil Pérez, M., Mlakar, I., Alcaraz Calero, J. M., García Clemente, F. J., & Martínez
Pérez, G. (2019). A Management Platform for Citizen’s Data Protection Regulation. In G. Wang, A. El
Saddik, X. Lai, G. Martinez Perez, & K. K. Choo (Eds.), Smart City and Informatization. iSCI 2019.
Communications in Computer and Information Science (Vol. 1122). Springer. doi:10.1007/978-981-
15-1301-5_6
Huertas Celdrán, A., Gil Pérez, M., Mlakar, I., Alcaraz Calero, J. M., García Clemente, F. J., Martínez
Pérez, G., & Bhuiyan, Z. A. (2020). PROTECTOR: Towards the protection of sensitive data in Europe
and the US. Computer Networks, 181, 107448. doi:10.1016/j.comnet.2020.107448
Ienca, M., & Haselager, P. (2016). Hacking the brain: Brain–computer interfacing technology and the eth-
ics of neurosecurity. Ethics and Information Technology, 18(2), 117–129. doi:10.100710676-016-9398-9
Ienca, M., Haselager, P., & Emanuel, E. (2018). Brain leaks and consumer neurotechnology. Nature
Biotechnology, 36(9), 805–810. doi:10.1038/nbt.4240 PMID:30188521
Khabarova, E. A., Denisova, N. P., Dmitriev, A. B., Slavin, K. V., & Verhagen Metman, L. (2018). Deep
Brain, Stimulation of the Subthalamic Nucleus in Patients with Parkinson Disease with Prior Pallidotomy
or Thalamotomy. Brain Sciences, 8(4), 66. doi:10.3390/brainsci8040066 PMID:29659494
Kubanek, J. (2018). Neuromodulation with transcranial focused ultrasound. Neurosurgical Focus FOC,
44(2), E14. doi:10.3171/2017.11.FOCUS17621 PMID:29385924
256
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Landau, O., Puzis, R., & Nissum, N. (2020). Mind Your Mind: EEG-Based Brain-Computer Interfaces
and Their Security in Cyber Space. ACM Computing Surveys, 53(1), 1–38. doi:10.1145/3372043
Lebedev, M. A., & Nicolelis, A. L. (2017). Brain-Machine Interfaces: From Basic Science to Neuropros-
theses and Neurorehabilitation. Physiological Reviews, 97(2), 767–837. doi:10.1152/physrev.00027.2016
PMID:28275048
Lee, J., Mok, E., Huang, J., Cui, L., Lee, A., Leung, V., Mercier, P., Shellhammer, S., Larson, L., As-
beck, P., Rao, R., Song, Y., Nurmikko, A., & Laiwalla, F. (2019) An Implantable Wireless Network of
Distributed Microscale Sensors for Neural Applications. 2019 9th International IEEE/EMBS Conference
on Neural Engineering (NER) (pp. 871-874). 10.1109/NER.2019.8717023
Legon, W., Bansal, P., Tyshynsky, R., Ai, L., & Mueller, J. K. (2018). Transcranial focused ultrasound
neuromodulation of the human primary motor cortex. Scientific Reports, 8(1), 10007. doi:10.103841598-
018-28320-1 PMID:29968768
Li, Q., Ding, D., & Conti, M. (2015). Brain-Computer Interface applications: Security and privacy chal-
lenges. 2015 IEEE Conference on Communications and Network Security (CNS), 663-666. 10.1109/
CNS.2015.7346884
López Bernal, S., Huertas Celdrán, A., Fernández Maimó, L., Martínez Pérez, G., Barros, M. T., &
Balasubramaniam, S. (2020). Cyberattacks on Miniature Brain Implants to Disrupt Spontaneous Neural
Signaling. IEEE Access: Practical Innovations, Open Solutions, 8, 152204–152222. doi:10.1109/AC-
CESS.2020.3017394
López Bernal, S., Huertas Celdrán, A., & Martínez Pérez, G. (2019). Cybersecurity on Brain-Computer
Interfaces: attacks and countermeasures. V Jornadas Nacionales de Investigación en Ciberseguridad
(JNIC2019), 198-199.
López Bernal, S., Huertas Celdrán, A., Martínez Pérez, G., Barros, M. T., & Balasubramaniam, S. (2020).
Security in Brain-Computer Interfaces: State-of-the-Art, Opportunities, and Future Challenges. ACM
Computing Surveys, 54(1), 35. doi:10.1145/3427376
Martinovic, I., Davies, D., & Frank, M. (2012) On the feasibility of side-channel attacks with brain-
computer interfaces. Proceedings of the 21st USENIX Security Symposium, 143-158.
McMahon, M., & Schukat, M. (2018). A low-Cost, Open-Source, BCI- VR Game Control Develop-
ment Environment Prototype for Game Based Neurorehabilitation. IEEE Games, Entertainment, Media
Conference (GEM), 1-9. 10.1109/GEM.2018.8516468
Musk, E. (2019). Neuralink An Integrated Brain-Machine Interface Platform With Thousands of Chan-
nels. Journal of Medical Internet Research, 21(10), e16194. doi:10.2196/16194 PMID:31642810
Neely, R. M., Piech, D. K., Santacruz, S. R., Maharbiz, M. M., & Carmena, J. M. (2018). Recent ad-
vances in neural dust: Towards a neural interface platform. Current Opinion in Neurobiology, 50, 64–71.
doi:10.1016/j.conb.2017.12.010 PMID:29331738
NeuroSky. (2019). http://neurosky.com
257
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Pais-Vieira, M., Lebedev, M., Kunicki, C., Wang, J., & Nicolelis, M. A. L. (2013). A Brain-to-Brain Interface
for Real-Time Sharing of Sensorimotor Information. Scientific Reports, 3(1), 1319. doi:10.1038rep01319
PMID:23448946
Pycroft, L., & Aziz, T. Z. (2018). Security of implantable medical devices with wireless connections:
The dangers of cyber-attacks. Expert Review of Medical Devices, 15(6), 403–406. doi:10.1080/174344
40.2018.1483235 PMID:29860880
Pycroft, L., Boccard, S. G., Owen, S. L. F., Stein, J. F., Fitzgerald, J. J., Green, A. L., & Aziz, T. Z.
(2016). Brainjacking: Implant Security Issues in Invasive Neuromodulation. World Neurosurgery, 92,
454–462. doi:10.1016/j.wneu.2016.05.010 PMID:27184896
Ramadan, R. A., & Vasilakos, A. V. (2017). Brain computer interface: Control signals review. Neuro-
computing, 223, 26–44. doi:10.1016/j.neucom.2016.10.024
Rao, R. P. N. (2019). Towards neural co-processors for the brain: Combining decoding and encoding in
brain–computer interfaces. Current Opinion in Neurobiology, 55, 142–151. doi:10.1016/j.conb.2019.03.008
PMID:30954862
Saboor, A., Gembler, F., Benda, M., Stawicki, P., Rezeika, A., Grichnik, R., & Volosyak, I. (2018). A
Browser-Driven SSVEP-Based BCI Web Speller. 2018 IEEE International Conference on Systems, Man,
and Cybernetics (SMC), 625–630. 10.1109/SMC.2018.00115
Sempreboni, D., & Viganò, L. (2018). Privacy, Security and Trust in the Internet of Neurons. arXiv.
Seo., D., Carmena, J. M., Rabaey, J. M., Alon, E., & Maharbiz, M. M. (2013). Neural Dust: An Ultra-
sonic, Low Power Solution for Chronic Brain-Machine Interfaces. arXiv.
Sowndhararajan, K., Minju, K., Ponnuvel, D., Se, P., & Songmun, K. (2018). Application of the P300
Event-Related Potential in the Diagnosis of Epilepsy Disorder: A Review. Scientia Pharmaceutica,
86(2), 10. doi:10.3390cipharm86020010 PMID:29587468
StatCounter Global Stats. (2020). Mobile & Tablet Android Version Market Share Worldwide. https://
gs.statcounter.com/android-version-market-share/mobile-tablet/worldwide
Sundararajan, K. (2017). Privacy and security issues in Brain Computer Interface [Unpublished master’s
thesis]. Auckland University of Technology, Auckland, New Zealand.
Takabi, H. (2016) Firewall for brain: Towards a privacy preserving ecosystem for BCI applications. IEEE
Conference on Communications and Network Security (CNS), 370-371. 10.1109/CNS.2016.7860516
Takabi, H., Bhalotiya, A., & Alohaly, M. (2016). Brain Computer Interface (BCI) Applications: Privacy
Threats and Countermeasures. 2016 IEEE 2nd International Conference on Collaboration and Internet
Computing (CIC), 102-111. 10.1109/CIC.2016.026
Tyler, W. J., Sanguinetti, J. L., Fini, M., & Hool, N. (2017). Non-invasive neural stimulation. Micro- and
Nanotechnology Sensors. Systems, and Applications, IX, 280–290. doi:10.1117/12.2263175
258
Cybersecurity Risks Associated With Brain-Computer Interface Classifications
Vaid, S., Singh, P., & Kaur, C. (2015). EEG Signal Analysis for BCI Interface: A Review. 2015 Fifth
International Conference on Advanced Computing & Communication Technologies, 143-147. 10.1109/
ACCT.2015.72
Waldert, S. (2016). Invasive vs. Non-Invasive Neuronal Signals for Brain-Machine Interfaces: Will One
Prevail? Frontiers in Neuroscience, 10, 295. doi:10.3389/fnins.2016.00295 PMID:27445666
Wirdatmadja, S. A., Barros, M. T., Koucheryavy, Y., Jornet, J. M., & Balasubramaniam, S. (2017).
Wireless Optogenetic Nanonetworks for Brain Stimulation: Device Model and Charging Protocols. IEEE
Transactions on Nanobioscience, 16(8), 859–872. doi:10.1109/TNB.2017.2781150 PMID:29364130
Yang, T., Hakimian, S., & Schwartz, T. H. (2014). Intraoperative ElectroCorticoGraphy (ECog): Indi-
cations, techniques, and utility in epilepsy surgery. Epileptic Disorders, 16(3), 271–279. doi:10.1684/
epd.2014.0675 PMID:25204010
Zander, T. O., Kothe, C., Jatzev, S., & Gaertner, M. (2010). Enhancing Human-Computer Interaction
with Input from Active and Passive Brain-Computer Interfaces. In D. Tan & A. Nijholt (Eds.), Brain-
Computer Interfaces. Human-Computer Interaction Series. Springer. doi:10.1007/978-1-84996-272-8_11
Zheng, H., Zhang, Z., Jiang, S., Yan, B., Shi, X., Xie, Y., Huang, X., Yu, Z., Liu, H., Weng, S., Nurmikko,
A., Zhang, Y., Peng, H., Xu, W., & Zhang, J. (2019). A shape-memory and spiral light-emitting device for
precise multisite stimulation of nerve bundles. Nature Communications, 10(1), 2790. doi:10.103841467-
019-10418-3 PMID:31243276
259
260
Compilation of References
Pande, A. S., & Thool, R. C. (2016, September). Survey on logical key hierarchy for secure group communication. 2016
International.
Sherman, A. T., & McGrew, D. A. (2003). Key establishment in large dynamic groups using one-way function trees.
IEEE Transactions on Software Engineering, 29(5), 444–458. doi:10.1109/TSE.2003.1199073
Kumar, N. S., & Lavanya, S. (2015). A novel scheme for secure group communication in multicast network. International
Journal of Security and Networks, 10(2), 65–75. doi:10.1504/IJSN.2015.070409
Bao, X., Liu, J., She, L., & Zhang, S. (2014, June). A key management scheme based on grouping within cluster. In
Proceeding of the 11th World Congress on Intelligent Control and Automation (pp. 3455- 3460). IEEE.
Kumar, S., Purusothaman, T. N. M., & Lavanya, S. (2013). Design and performance analysis of scalable and efficient group
key Manage- ment scheme [SEGKMS] for group communication in multicast net- works. Life Science Journal, 10(2).
Albakri, A., Harn, L., & Song, S. (2019). Hierarchical key manage- ment scheme with probabilistic security in a wireless
sensor network (WSN). Security and Communication Networks, 2019.
Vijayakumar, P., Bose, S., Kannan, A., & Jegatha Deborah, L. (2013). Computation and Communication Efficient Key
Distribution. Academic Press.
Zhu, S., & Jajodia, S. (2003, December). Scalable group rekeying for secure multicast: A survey. In International Work-
shop on Distributed Computing (pp. 1-10). Springer. 10.1007/978-3-540-24604-6_1
Eltoweissy, M., Heydari, M. H., Morales, L., & Sudborough, I. H. (2004). Combinatorial optimization of group key
management. Journal of Network and Systems Management, 12(1), 33–50. doi:10.1023/B:JONS.0000015697.38671.ec
Wong, C. K., Gouda, M., & Lam, S. S. (2000). Secure group communications using key graphs. IEEE/ACM Transactions
on Networking, 8(1), 16–30. doi:10.1109/90.836475
He, X., Niedermeier, M., & De Meer, H. (2013). Dynamic key management in wireless sensor networks: A survey.
Journal of Network and Computer Applications, 36(2), 611–622. doi:10.1016/j.jnca.2012.12.010
Dave, M. (2016, March). Storage as a parameter for classifying dynamic key management schemes proposed for WSNs.
In 2016 International Conference on Computational Techniques in Information and Communication Technologies
(ICCTICT) (pp. 51-56). IEEE.
Eschenauer, L., & Gligor, V. D. (2002). A key-management scheme for distributed sensor networks. Proceedings of the
9th ACM Conference on Computer and Communication Security, 41–47. 10.1145/586110.586117
Chen, C. T., Huang, S. Y., & Lin, I. C. (2012). Providing perfect forward secrecy for location-aware wireless sensor net-
works. EURASIP Journal on Wireless Communications and Networking, 2012(1), 241. doi:10.1186/1687-1499-2012-241
Compilation of References
Messai, M. L., Aliouat, M., & Seba, H. (2010). Tree based scheme for key Management in wireless sensor networks.
EURASIP Journal on Wireless Communications and Networking, 2010(1), 1–10. doi:10.1155/2010/910695
Liu, D., & Ning, P. (2003, October). Location-based pairwise key establishments for static sensor networks. In Proceed-
ings of the 1st ACM workshop on Security of ad hoc and sensor networks (pp. 72-82). 10.1145/986858.986869
Eltoweissy, M., Moharrum, M., & Mukkamala, R. (2006). Dynamic key management in sensor networks. IEEE Com-
munications Magazine, 44(4), 122–130. doi:10.1109/MCOM.2006.1632659
Dutertre, B., Cheung, S., & Levy, J. (2004). Lightweight key management in wireless sensor networks by leveraging
initial trust. Technical Report SRI-SDL-04-02, SRI International.
Du, W., Deng, J., Han, Y. S., Chen, S., & Varshney, P. K. (2004, March). A key management scheme for wireless sensor
networks using deployment knowledge. In IEEE INFOCOM 2004 (Vol. 1). IEEE.
Xu, Y., & Sun, Y. (2005, December). A new group rekeying method in secure multicast. In International Conference on
Computational and Information Science (pp. 155-160). Springer. 10.1007/11596981_23
Wallner, D., Harder, E., & Agee, R. (1999). Key management for multicast: Issues and architectures. RFC 2627.
Yousefpoor, M. S., & Barati, H. (2019). Dynamic key management algorithms in wireless sensor networks: A survey.
Computer Communications, 134, 52–69. doi:10.1016/j.comcom.2018.11.005
Seetha, R., & Saravanan, R. (2015). A survey on group key management schemes. Cybernetics and Information Tech-
nologies, 15(3), 3–25. doi:10.1515/cait-2015-0038
Chen, C. Y., & Chao, H. C. (2014). A survey of key distribution in wireless sensor networks. Security and Communica-
tion Networks, 7(12), 2495–2508. doi:10.1002ec.354
Barskar, R., & Chawla, M. (2016). A survey on efficient group key management schemes in wireless networks. Indian
Journal of Science and Technology, 9(14), 1–16. doi:10.17485/ijst/2016/v9i14/87972
Rahman, M., & Sampalli, S. (2015). An efficient pairwise and group key management protocol for wireless sensor
network. Wireless Personal Communications, 84(3), 2035–2053. doi:10.100711277-015-2546-4
netlab. (2017). Warning: Satori, a Mirai Branch Is Spreading in Worm Style on Port 37215 and 52869. Retrieved from
https://blog.netlab.360.com/warning-satori-a-new-mirai-variant-is-spreading-in-worm-style-on-port-37215-and-52869-
en/
Jain, A. K., & Gupta, B. B. (2019). A machine learning based approach for phishing detection using hyperlinks infor-
mation. Journal of Ambient Intelligence and Humanized Computing, 10(5), 2015–2028. doi:10.100712652-018-0798-z
Ouaguid, A., Abghour, N., & Ouzzif, M. (2018). A novel security framework for managing android permissions us-
ing blockchain technology. International Journal of Cloud Applications and Computing, 8(1), 55–79. doi:10.4018/
IJCAC.2018010103
Stergiou, C., Psannis, K. E., Gupta, B. B., & Ishibashi, Y. (2018). Security, privacy & efficiency of sustainable cloud
computing for big data & IoT. Sustainable Computing: Informatics and Systems, 19, 174–184. doi:10.1016/j.sus-
com.2018.06.003
Zheng, Q., Wang, X., Khan, M. K., Zhang, W., Gupta, B. B., & Guo, W. (2017). A lightweight authenticated encryp-
tion scheme based on chaotic scml for railway cloud service. IEEE Access: Practical Innovations, Open Solutions, 6,
711–722. doi:10.1109/ACCESS.2017.2775038
261
Compilation of References
Rafaeli, S., & Hutchison, D. (2003). A survey of key management for secure group communication. ACM Computing
Surveys, 35(3), 309–329. doi:10.1145/937503.937506
Ng, W. H. D., Howarth, M., Sun, Z., & Cruickshank, H. (2007). Dynamic balanced key tree management for secure
multicast communications. IEEE Transactions on Computers, 56(5), 590–605. doi:10.1109/TC.2007.1022
Ng, W. H. D., Cruickshank, H., & Sun, Z. ((2006) Scalable balanced batch rekeying for secure group communication.
Computers & Security, 25(4), 265-273.
Tiloca, M., Dini, G., Rizki, K., & Raza, S. (2019). Group rekeying based on member join history. International Journal
of Information Security, 1–39. doi:10.100710207-019-00451-0
Howarth, M. P., Iyengar, S., Sun, Z., & Cruickshank, H. (2004). Dynamics of key management in secure satellite multicast.
IEEE Journal on Selected Areas in Communications, 22(2), 308–319. doi:10.1109/JSAC.2003.819978
Bilal, M., & Kang, S. G. (2017). A secure key agreement protocol for dynamic group. Cluster Computing, 20(3),
2779–2792. doi:10.100710586-017-0853-0
Duma, C., Shahmehri, N., & Lambrix, P. (2003, June). A hybrid key tree scheme for multicast to balance security and
efficiency requirements. In WET ICE 2003. Proceedings. Twelfth IEEE International Workshops on Enabling Technolo-
gies: Infrastructure for Collaborative Enterprises, 2003 (pp. 208-213). IEEE.
Ab Malek, M. S. B., Ahmadon, M. A. B., Yamaguchi, S., & Gupta, B. B. (2016, October). On privacy verification in the
IoT service based on PN 2. In 2016 IEEE 5th Global Conference on Consumer Electronics (pp. 1-4). IEEE.
Adat, V., & Gupta, B. B. (2018). Security in Internet of Things: Issues, challenges, taxonomy, and architecture. Telecom-
munication Systems, 67(3), 423–441. doi:10.100711235-017-0345-9
Adepu, S., Kandasamy, N. K., & Mathur, A. (2018). Epic: An electric power testbed for research and training in cyber
physical systems security. In Computer Security (pp. 37–52). Springer.
Aghajari, E., & Chandrashekhar, G. D. (2017). Self-organizing map based extended fuzzy C-means (SEEFC) algorithm
for image segmentation. Applied Soft Computing, 54, 347–363. doi:10.1016/j.asoc.2017.01.003
Agrawal, D. P., & Wang, H. (2018). Computer and Cyber Security: Principles, Algorithm, Applications, and Perspectives
(B. B. Gupta, Ed.; 1st ed.). CRC Press. doi:10.1201/9780429424878
Ahlawat, P., & Dave, M. (2018). An attack resistant key predistribution scheme for wireless sensor networks. Journal
of King Saud University-Computer and Information Sciences.
Ahlawat, P., & Dave, M. (2018). An attack model based highly secure key management scheme for wireless sensor
networks. Procedia Computer Science, 125, 201–207. doi:10.1016/j.procs.2017.12.028
Ahmadian, M. M., Shajari, M., & Shafiee, M. A. (2020). Industrial control system security taxonomic framework with
application to a comprehensive incidents survey. International Journal of Critical Infrastructure Protection, 29, 100356.
doi:10.1016/j.ijcip.2020.100356
Ahmadi, M., Ulyanov, D., Semenov, S., Trofimov, M., & Giacinto, G. (2016, March). Novel feature extraction, selection
and fusion for effective malware family classification. In Proceedings of the sixth ACM conference on data and applica-
tion security and privacy (pp. 183-194). ACM.
262
Compilation of References
Ahmadlou, M., Al‐Fugara, A. K., Al‐Shabeeb, A. R., Arora, A., Al‐Adamat, R., Pham, Q. B., Al-Ansari, N., Linh, N.
T. T., & Sajedi, H. (2021). Flood susceptibility mapping and assessment using a novel deep learning model combining
multilayer perceptron and autoencoder neural networks. Journal of Flood Risk Management, 14(1), e12683. doi:10.1111/
jfr3.12683
Ahn, M., Lee, M., Choi, J., & Jun, S. (2014). A Review of Brain-Computer Interface Games and an Opinion Survey from
Researchers, Developers, and Users. Sensors (Basel), 14(8), 14601–14633. doi:10.3390140814601 PMID:25116904
Airehrour, D., Gutierrez, J., & Ray, S. K. (2017). A trust-aware RPL routing protocol to detect blackhole and selective
forwarding attacks. Journal of Telecommunications and the Digital Economy, 5(1), 50–69.
Ajrawi, S., Rao, R., & Sarkar, M. (2021). Cybersecurity in Brain-Computer Interfaces: RFID-based design-theoretical
framework. Informatics in Medicine Unlocked, 22, 100489. doi:10.1016/j.imu.2020.100489
Al Shuhaimi, F., Jose, M., & Singh, A. V. (2016, September). Software defined network as solution to overcome security
challenges in IoT. In 2016 5th International Conference on Reliability, Infocom Technologies and Optimization (Trends
and Future Directions)(ICRITO) (pp. 491-496). IEEE.
Alaba, F. A., Othman, M., Hashem, I. A. T., & Alotaibi, F. (2017). Internet of Things security: A survey. Journal of
Network and Computer Applications, 88, 10–28. doi:10.1016/j.jnca.2017.04.002
Aldowah, H., Rehman, S. U., & Umar, I. (2018, June). Security in internet of things: issues, challenges and solutions. In
International Conference of Reliable Information and Communication Technology (pp. 396-405). Springer.
Al-Garadi, M. A., Mohamed, A., Al-Ali, A. K., Du, X., Ali, I., & Guizani, M. (2020). A survey of machine and deep
learning methods for internet of things (IoT) security. IEEE Communications Surveys and Tutorials, 22(3), 1646–1685.
Alghamdi, T. A. (2019). Convolutional technique for enhancing security in wireless sensor networks against malicious
nodes. Human-centric Computing and Information Sciences, 9(1), 38. doi:10.118613673-019-0198-1
Ali, I., Sabir, S., & Ullah, Z. (2019). Internet of things security, device authentication and access control: a review.
arXiv preprint arXiv:1901.07309.
Almogren, A. S. (2020). Intrusion detection in Edge-of-Things computing. Journal of Parallel and Distributed Comput-
ing, 137, 259–265. doi:10.1016/j.jpdc.2019.12.008
Al-Nuaimi, F. A., Al-Nuaimi, R. J., Al-Dhaheri, S. S., Ouhbi, S., & Belkacem, A. N. (2020). Mind Drone Chasing Using
EEG-based Brain Computer Interface. 2020 16th International Conference on Intelligent Environments (IE), 1. 10.1109/
IE49459.2020.9154926
Alomari, E., Manickam, S., Gupta, B. B., Anbar, M., Saad, R. M., & Alsaleem, S. (2016). A survey of botnet-based
ddos flooding attacks of application layer: Detection and mitigation approaches. In Handbook of research on modern
cryptographic solutions for computer and cyber security (pp. 52–79). IGI Global.
Al-Qerem, A., Alauthman, M., Almomani, A., & Gupta, B. B. (2020). IoT transaction processing through cooperative
concurrency control on fog–cloud computing environment. Soft Computing, 24(8), 5695–5711.
Alrawais, A., Alhothaily, A., Hu, C., & Cheng, X. (2017). Fog computing for the internet of things: Security and privacy
issues. IEEE Internet Computing, 21(2), 34–42.
Altaweel, M., Bone, C., & Abrams, J. (2019). Documents as data: A content analysis and topic modeling approach for
analyzing responses to ecological disturbances. Ecological Informatics, 51, 82–95. doi:10.1016/j.ecoinf.2019.02.014
263
Compilation of References
Aly, M., Khomh, F., Haoues, M., Quintero, A., & Yacout, S. (2019). Enforcing security in Internet of Things frameworks:
A systematic literature review. Internet of Things, 6, 100050. doi:10.1016/j.iot.2019.100050
Ammar, M., Russello, G., & Crispo, B. (2018). Internet of Things: A survey on the security of IoT frameworks. Journal
of Information Security and Applications, 38, 8–27.
Amofa, S., Sifah, E. B., Kwame, O. B., Abla, S., Xia, Q., Gee, J. C., & Gao, J. (2018, September). A blockchain-based
architecture framework for secure sharing of personal health data. In 2018 IEEE 20th International Conference on e-
Health Networking, Applications and Services (Healthcom) (pp. 1-6). IEEE. 10.1109/HealthCom.2018.8531160
An, H., Kim, J., & Lee, S. (2016). Design of an asynchronous brain-computer interface for control of a virtual Avatar.
2016 4th International Winter Conference on Brain-Computer Interface (BCI) (pp. 1-2). 10.1109/IWW-BCI.2016.7457463
Ande, R., Adebisi, B., Hammoudeh, M., & Saleem, J. (2020). Internet of Things: Evolution and technologies from a
security perspective. Sustainable Cities and Society, 54, 101728.
Andrea, I., Chrysostomou, C., & Hadjichristofi, G. (2015, July). Internet of Things: Security vulnerabilities and chal-
lenges. In 2015 IEEE symposium on computers and communication (ISCC) (pp. 180-187). IEEE.
Anusha, K., & Sathiyamoorthy, E. (2016). A decision tree-based rule formation with combined PSO-GA algorithm
for intrusion detection system. International Journal of Internet Technology and Secured Transactions, 6(3), 186–202.
doi:10.1504/IJITST.2016.080399
Arghire, I. (2017). Mirai Variant Has Bitcoin Mining Capabilities. Retrieved from https://www.securityweek.com/mirai-
variant-has-bitcoin-mining-capabilities
Arias, O., Wurm, J., Hoang, K., & Jin, Y. (2015). Privacy and security in internet of things and wearable devices. IEEE
Transactions on Multi-Scale Computing Systems, 1(2), 99–109.
Aricò, P., Borghini, G., Di Flumeri, G., Sciaraffa, N., & Babiloni, F. (2018). Passive BCI beyond the lab: Current trends
and future directions. Physiological Measurement, 39(8), 08TR02. Advance online publication. doi:10.1088/1361-6579/
aad57e PMID:30039806
Arıs, A., Oktug, S. F., & Voigt, T. (2018). Security of internet of things for a reliable internet of services. Academic Press.
Arrichiello, F., Di Lillo, P., Di Vito, D., Antonelli, G., & Chiaverini, S. (2017). Assistive robot operated via P300-
based brain computer interface. 2017 IEEE International Conference on Robotics and Automation (ICRA), 1. 10.1109/
ICRA.2017.7989714
Arzamendi, P., Bing, M., & Soluk, K. (2018). The ARC of Satori. Retrieved from https://www.netscout.com/blog/asert/
arc-satori
ASERT Team. (2018). OMG - Mirai Minions are Wicked. Retrieved from https://www.netscout.com/blog/asert/omg-
mirai-minions-are-wicked
Ashfahani, A., Pratama, M., Lughofer, E., & Ong, Y. S. (2020). DEVDAN: Deep evolving denoising autoencoder. Neu-
rocomputing, 390, 297–314. doi:10.1016/j.neucom.2019.07.106
Athiwaratkun, B., & Stokes, J. W. (2017, March). Malware classification with LSTM and GRU language models and
a character-level CNN. In 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
(pp. 2482-2486). IEEE.
Aydemir, O. (2015). Improving classification accuracy of EEG based brain computer interface signals. 2015 23nd Signal
Processing and Communications Applications Conference (SIU), 1. 10.1109/SIU.2015.7130442
264
Compilation of References
Badcock, N. A., Mousikou, P., Mahajan, Y., de Lissa, P., Thie, J., & McArthur, G. (2013). Validation of the Emotiv EPOC®
EEG gaming system for measuring research quality auditory ERPs. PeerJ, 1, e38. doi:10.7717/peerj.38 PMID:23638374
Bailey, M., Cooke, E., Jahanian, F., Xu, Y., & Karir, M. (2009). A Survey of Botnet Technology and Defenses. In Proc.
of the 2009 Cybersecurity Applications & Technology Conference for Homeland Security (pp.299-304). Academic Press.
Ballano, M. (2015). Is there an Internet-of-Things vigilante out there? Retrieved from https://www.symantec.com/con-
nect/blogs/there-internet-things-vigilante-out-there
Beaver, J. M., Borges-Hink, R. C., & Buckner, M. A. (2013, December). An evaluation of machine learning methods to
detect malicious SCADA communications. In 2013 12th international conference on machine learning and applications
(Vol. 2, pp. 54-59). IEEE. 10.1109/ICMLA.2013.105
Belavagi, M. C., & Muniyal, B. (2016). Performance evaluation of supervised machine learning algorithms for intrusion
detection. Procedia Computer Science, 89, 117–123. doi:10.1016/j.procs.2016.06.016
Bentabet, N., & Berrached, N. (2016). Synchronous P300 based BCI to control home appliances. 2016 8th International
Conference on Modelling, Identification and Control (ICMIC) (pp. 835-838). 10.1109/ICMIC.2016.7804230
Berman, D. S., Buczak, A. L., Chavis, J. S., & Corbett, C. L. (2019). A survey of deep learning methods for cyber se-
curity. Information (Basel), 10(4), 122. doi:10.3390/info10040122
Bernard, Y., Hueber, N., & Girau, B. (2020, September). A fast algorithm to find Best Matching Units in Self-Organizing
Maps. In International Conference on Artificial Neural Networks (pp. 825-837). Springer.
Bertino, E., & Islam, N. (2017). Botnets and internet of things security. Computer, 50(2), 76–79.
Bezerra, V. H., da Costa, V. G. T., Barbon, J., Miani, R. S., & Zarpelão, B. B. (2019). IoTDS: A One-Class Classification
Approach to Detect Botnets in Internet of Things Devices. Sensors (Basel), 19(14), 3188.
Bharathi, Tanguturi, JayaKumar, & Selvamani. (2012). Node capture attack in Wireless Sensor Network:A survey. 2012
IEEE International Conference on Computational Intelligence and Computing Research, ICCIC 2012, 1-3. 10.1109/
ICCIC.2012.6510237
Bhaskaran, K., Ilfrich, P., Liffman, D., Vecchiola, C., Jayachandran, P., Kumar, A., ... Teo, E. G. (2018, April). Double-
blind consent-driven data sharing on blockchain. In 2018 IEEE International Conference on Cloud Engineering (IC2E)
(pp. 385-391). IEEE. 10.1109/IC2E.2018.00073
Bhattarai, S., & Wang, Y. (2018). End-to-end trust and security for Internet of Things applications. Computer, 51(4),
20–27. doi:10.1109/MC.2018.2141038
Bhatt, R., Maheshwary, P., Shukla, P., Shukla, P., Shrivastava, M., & Changlani, S. (2020). Implementation of Fruit Fly
Optimization Algorithm (FFOA) to escalate the attacking efficiency of node capture attack in Wireless Sensor Networks
(WSN). Computer Communications, 149, 134–145. doi:10.1016/j.comcom.2019.09.007
Bhatt, S., & Ragiri, P. R. (2021). Security trends in Internet of Things: A survey. SN Applied Sciences, 3(1), 1–14.
Bhushan, K., & Gupta, B. B. (2019). Distributed denial of service (DDoS) attack mitigation in software defined network
(SDN)-based cloud computing environment. Journal of Ambient Intelligence and Humanized Computing, 10(5), 1985–1997.
265
Compilation of References
Bikson, M., Brunoni, A. R., Charvet, L. E., Clark, V. P., Cohen, L. G., Deng, Z. D., Dmochowski, J., Edwards, D.
J., Frohlich, F., Kappenman, E. S., Lim, K. O., Loo, C., Mantovani, A., McMullen, D. P., Parra, L. C., Pearson, M.,
Richardson, J. D., Rumsey, J. M., Sehatpour, P., ... Losanby, S. H. (2018). Rigor and reproducibility in research with
transcranial electrical stimulation: An NIMH-sponsored workshop. Brain Stimulation, 11(3), 465–480. doi:10.1016/j.
brs.2017.12.008 PMID:29398575
Binbusayyis, A., & Vaiyapuri, T. (2020). Comprehensive analysis and recommendation of feature evaluation measures
for intrusion detection. Heliyon, 6(7), e04262. doi:10.1016/j.heliyon.2020.e04262 PMID:32685709
Binnendijk, A., Marler, T., & Bartels, E. M. (2020). Brain-Computer Interfaces: U.S. Military Applications and Implica-
tions, An Initial Assessment. RAND Corporation. doi:10.7249/RR2996
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. The Journal of Machine Learning Research,
3, 993-1022.
Bonaci, T., Bushnell, L., & Poovendran, R. (2010, December). Node capture attacks in wireless sensor networks: A
system theoretic approach. In 49th IEEE Conference on Decision and Control (CDC) (pp. 6765-6772). IEEE. 10.1109/
CDC.2010.5717499
Bonaci, T., Calo, R., & Chizeck, H. J. (2015). App Stores for the Brain: Privacy and Security in Brain-Computer Inter-
faces. IEEE Technology and Society Magazine, 34(2), 32–39. doi:10.1109/MTS.2015.2425551
Bonaci, T., Herron, J., Matlack, C., & Chizeck, H. J. (2015). Securing the Exocortex: A Twenty-First Century Cybernetics
Challenge. IEEE Technology and Society Magazine, 34(3), 44–51. doi:10.1109/MTS.2015.2461152
Bonderud, D. (2016). Leaked Mirai Malware Boosts IoT Insecurity Threat Level. Retrieved from https://securityintel-
ligence.com/news/leaked-mirai-malware-boosts-iot-insecurity-threat-level/
Bormann, C., Castellani, A. P., & Shelby, Z. (2012). Coap: An application protocol for billions of tiny internet nodes.
IEEE Internet Computing, 16(2), 62–67.
Boyer, S. A. (1999). SCADA: Supervisory control and data acquisition (Vol. 3). ISA.
Boyes, H., Hallaq, B., Cunningham, J., & Watson, T. (2018). The industrial internet of things (IIoT): An analysis frame-
work. Computers in Industry, 101, 1–12. doi:10.1016/j.compind.2018.04.015
Brakerski, Z., & Vaikuntanathan, V. (2014). Efficient fully homomorphic encryption from (standard) LWE. SIAM Journal
on Computing, 43(2), 831–871.
Burnap, P., French, R., Turner, F., & Jones, K. (2018). Malware classification using self organising feature maps and
machine activity data. Computers & Security, 73, 399–410. doi:10.1016/j.cose.2017.11.016
Butun, I., Österberg, P., & Song, H. (2019). Security of the Internet of Things: Vulnerabilities, attacks, and countermea-
sures. IEEE Communications Surveys and Tutorials, 22(1), 616–644.
Cabău, G., Buhu, M., & Oprişa, C. (2016). Malware classification using filesystem footprints. 2016 IEEE International
Conference on Automation, Quality and Testing, Robotics (AQTR), 1-6. 10.1109/AQTR.2016.7501294
Campbell, A., & Wu, C. (2018). Chronically Implanted Intracranial Electrodes: Tissue Reaction and Electrical Changes.
Micromachines, 9(9), 430. doi:10.3390/mi9090430 PMID:30424363
Canfora, G., Mercaldo, F., Visaggio, C. A., & Di Notte, P. (2014). Metamorphic malware detection using code metrics.
Information Security Journal: A Global Perspective, 23(3), 57-67.
266
Compilation of References
Cash, M., & Bassiouni, M. (2018, September). Two-tier permission-ed and permission-less blockchain for secure data
sharing. In 2018 IEEE International Conference on Smart Cloud (SmartCloud) (pp. 138-144). IEEE. 10.1109/Smart-
Cloud.2018.00031
Cech, H. L., Großmann, M., & Krieger, U. R. (2019, June). A fog computing architecture to share sensor data by means
of blockchain functionality. In 2019 IEEE International Conference on Fog Computing (ICFC) (pp. 31-40). IEEE.
10.1109/ICFC.2019.00013
Ceron, J. M., Steding-Jessen, K., Hoepers, C., Granville, L. Z., & Margi, C. B. (2019). Improving IoT Botnet Investiga-
tion Using an Adaptive Network Layer. Sensors (Basel), 19(3), 727.
Chahid, Y., Benabdellah, M., & Azizi, A. (2017, April). Internet of things security. In 2017 International Conference on
Wireless Technologies, Embedded and Intelligent Systems (WITS) (pp. 1-6). IEEE. doi:10.1109/ICOASE.2018.8548785
Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: A survey. ACM computing surveys (CSUR), 41(3),
1-58.
Chan, H., Perrig, A., & Song, D. (2003, May). Random key predistribution schemes for sensor networks. In 2003 Sym-
posium on Security and Privacy, 2003 (pp. 197-213). IEEE. 10.1109/SECPRI.2003.1199337
Chan, K., & Fekri, F. (2007, May). Node compromise attacks and network connectivity. In Defense Transformation and
Net-Centric Systems 2007 (Vol. 6578, p. 65780W). International Society for Optics and Photonics.
Chasaki, D., & Mansour, C. (2015). Security challenges in the internet of things. International Journal of Space-Based
and Situated Computing, 5(3), 141–149.
Chaudhary, P., & Agrawal, R. (2018). Emerging Threats to Security and Privacy in Brain Computer Interface Interna-
tional Journal of Advanced Studies of Scientific Research, 3(12). https://ssrn.com/abstract=3326692
Chaudhary, P., Gupta, B. B., Chang, X., Nedjah, N., & Chui, K. T. (2021). Enhancing big data security through integrating
XSS scanner into fog nodes for SMEs gain. Technological Forecasting and Social Change, 168, 120754. doi:10.1016/j.
techfore.2021.120754
Chaudhary, U., Xia, B., Silvoni, S., Cohen, L. G., & Birbaumer, N. (2017). Brain–Computer Interface–Based Communica-
tion in the Completely Locked-In State. PLoS Biology, 15(1), 1–25. doi:10.1371/journal.pbio.1002593 PMID:28141803
Chegini, H., & Mahanti, A. (2019, December). A Framework of Automation on Context-Aware Internet of Things (IoT)
Systems. In Proceedings of the 12th IEEE/ACM International Conference on Utility and Cloud Computing Companion
(pp. 157-162). IEEE.
Chen, Y. K. (2012, January). Challenges and opportunities of internet of things. In 17th Asia and South Pacific design
automation conference (pp. 383-388). IEEE.
Chen, H., Perrig, A., & Song, D. (2003) Random Key Predistribution Schemes for Sensor Networks. Proceedings of the
2003 IEEE symposium on security and privacy, 197–213.
Chen, K., Zhang, S., Li, Z., Zhang, Y., Deng, Q., Ray, S., & Jin, Y. (2018). Internet-of-things security and vulnerabilities:
Taxonomy, challenges, and practice. Journal of Hardware and Systems Security, 2(2), 97–110.
Chhabra, Gupta, & Almomani. (2013). A novel solution to handle DDOS attack in MANET. Academic Press.
Chhabra, M., Gupta, B., & Almomani, A. (2013). A novel solution to handle DDOS attack in MANET. Academic Press.
267
Compilation of References
Chhabra, M., Gupta, B., & Almomani, A. (2013). A novel solution to handle DDOS attack in MANET. Journal of
Information Security, 4(3), 34631. Advance online publication. doi:10.4236/jis.2013.43019
Chiappa, K. H., & Ropper, A. H. (1982). Evoked Potentials in Clinical Medicine. The New England Journal of Medicine,
306(19), 1140–1150. doi:10.1056/NEJM198205133061904 PMID:7040957
Chizeck, H. J., & Bonaci, T. (2014). Brain-Computer Interface Anonymizer (Patent application US20140228701A1).
United States Patent and Trademark Office.
Choi, S. (2020). Combined kNN Classification and hierarchical similarity hash for fast malware detection. Applied Sci-
ences, 10(15), 5173.
Choo, A., & May, A. (2014). Virtual mindfulness meditation: Virtual reality and electroencephalography for health
gamification. 2014 IEEE Games Media Entertainment, 1. doi:10.1109/GEM.2014.7048076
Chouhan, L., Chauhan, N., Mahapatra, A. S., & Agarwal, V. (2020). A survey on the applications of machine learning
in wireless sensor networks. International Journal of High Performance Computing and Networking, 16(4), 197–220.
doi:10.1504/IJHPCN.2020.113779
Chow, J. K., Su, Z., Wu, J., Tan, P. S., Mao, X., & Wang, Y. H. (2020). Anomaly detection of defects on concrete struc-
tures with the convolutional autoencoder. Advanced Engineering Informatics, 45, 101105. doi:10.1016/j.aei.2020.101105
Chui, K. T., Vasant, P., & Liu, R. W. (2019). Smart city is a safe city: information and communication technology–enhanced
urban space monitoring and surveillance systems: the promise and limitations. In Smart Cities: Issues and Challenges
(pp. 111-124). Elsevier. doi:10.1016/B978-0-12-816639-0.00007-7
Chui, K. T., Liu, R. W., Zhao, M., & De Pablos, P. O. (2020). Predicting students’ performance with school and family
tutoring using generative adversarial network-based deep support vector machine. IEEE Access: Practical Innovations,
Open Solutions, 8, 86745–86752. doi:10.1109/ACCESS.2020.2992869
Cimpanu, C. (2018). New Mirai Variant Focuses on Turning IoT Devices into Proxy Servers. Retrieved from https://www.
bleepingcomputer.com/news/security/new-mirai-variant-focuses-on-turning-iot-devices-into-proxy-servers/
Cisco. (2020). Cisco Annual Internet Report (2018–2023) White Paper. Retrieved from https://www.cisco.com/c/en/us/
solutions/collateral/executive-perspectives/annual-internet-report/white-paper-c11-741490.html
Conti, M., Dehghantanha, A., Franke, K., & Watson, S. (2018). Internet of Things security and forensics: Challenges
and opportunities. Academic Press.
Crea, S., Nann, M., Trigili, E., Cordella, F., Baldoni, A., Badesa, F. J., Catalán, J. M., Zollo, L., Vitiello, N., Aracil, N.
G., & Soekadar, S. R. (2018). Feasibility and safety of shared EEG/EOG and vision-guided autonomous whole-arm
exoskeleton control to perform activities of daily living. Scientific Reports, 8(1), 1. doi:10.103841598-018-29091-5
PMID:30018334
Croman, K., Decker, C., Eyal, I., Gencer, A. E., Juels, A., Kosba, A., & Song, D. (2016, February). On scaling decen-
tralized blockchains. In International conference on financial cryptography and data security (pp. 106-125). Springer.
Curran, M. T., Yang, J., Merrill, N., & Chuang, J. (2016). Passthoughts authentication with low cost EarEEG. 2016
38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 1979-1982.
10.1109/EMBC.2016.7591112
Cvitić, I., Peraković, D., Periša, M., & Gupta, B. (2021). Ensemble machine learning approach for classification of IoT
devices in smart home. International Journal of Machine Learning and Cybernetics, 1-24. https://www.ll.mit.edu/r-d/
datasets/1998-darpa-intrusion-detection-evaluation-dataset
268
Compilation of References
Cvitić, I., Peraković, D., Periša, M., & Gupta, B. (2021). Ensemble machine learning approach for classification of IoT
devices in smart home. International Journal of Machine Learning and Cybernetics, 1–24.
da Costa, K. A., Papa, J. P., Lisboa, C. O., Munoz, R., & de Albuquerque, V. H. C. (2019). Internet of Things: A survey
on machine learning-based intrusion detection approaches. Computer Networks, 151, 147–157. doi:10.1016/j.com-
net.2019.01.023
Dabbagh, M., & Rayes, A. (2019). Internet of things security and privacy. In Internet of Things from hype to reality (pp.
211–238). Springer.
Dadoun, A., & Troncy, R. (2020). Many-to-one Recurrent Neural Network for Session-based Recommendation. arXiv
preprint arXiv:2008.11136.
Dahiya, A., & Gupta, B. B. (2021). A reputation score policy and Bayesian game theory based incentivized mechanism
for DDoS attacks mitigation and cyber defense. Future Generation Computer Systems, 117, 193–204.
Dahl, G. E., Stokes, J. W., Deng, L., & Yu, D. (2013, May). Large-scale malware classification using random projec-
tions and neural networks. In 2013 IEEE International Conference on Acoustics, Speech and Signal Processing (pp.
3422-3426). IEEE.
Das, S., Koperski, M., Bremond, F., & Francesca, G. (2018, November). Deep-temporal lstm for daily living action
recognition. In 2018 15th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)
(pp. 1-6). IEEE. 10.1109/AVSS.2018.8639122
Das, A. K., Zeadally, S., & He, D. (2018). Taxonomy and analysis of security protocols for Internet of Things. Future
Generation Computer Systems, 89, 110–125.
David, O. E., & Netanyahu, N. S. (2015, July). Deepsign: Deep learning for automatic malware signature generation
and classification. In 2015 International Joint Conference on Neural Networks (IJCNN) (pp. 1-8). IEEE. 10.1109/
IJCNN.2015.7280815
Denning, T., Matsuoka, Y., & Kohno, T. (2009). Neurosecurity: Security and privacy for neural devices. Neurosurgical
Focus FOC, 27(1), E7. doi:10.3171/2009.4.FOCUS0985 PMID:19569895
De, P., Liu, Y., & Das, S. K. (2006, June). Modeling node compromise spread in wireless sensor networks using epidemic
theory. In 2006 International Symposium on a World of Wireless, Mobile and Multimedia Networks (WoWMoM’06).
IEEE. 10.1109/WOWMOM.2006.74
Devry, J. (2016). Mirai Botnet Infects Devices in 164 Countries. Retrieved from https://www.cybersecurity-insiders.com/
mirai-botnet-infects-devices-in-164-countries/
Dhillon, A., & Verma, G. K. (2020). Convolutional neural network: A review of models, methodologies and applications
to object detection. Progress in Artificial Intelligence, 9(2), 85–112. doi:10.100713748-019-00203-0
Diallo, B., Hu, J., Li, T., Khan, G. A., Liang, X., & Zhao, Y. (2021). Deep embedding clustering based on contractive
autoencoder. Neurocomputing, 433, 96–107. doi:10.1016/j.neucom.2020.12.094
Dib, O., Brousmiche, K. L., Durand, A., Thea, E., & Hamida, E. B. (2018). Consortium blockchains: Overview, applica-
tions and challenges. International Journal on Advances in Telecommunications, 11(1-2).
269
Compilation of References
Dinh, T. T. A., Wang, J., Chen, G., Liu, R., Ooi, B. C., & Tan, K. L. (2017, May). Blockbench: A framework for ana-
lyzing private blockchains. In Proceedings of the 2017 ACM International Conference on Management of Data (pp.
1085-1100). 10.1145/3035918.3064033
Donchin, E. (1979). Event-related Brain Potentials: A Tool in the Study of Human Information Processing. Evoked Brain
Potentials and Behavior, 13–88. doi:10.1007/978-1-4684-3462-0_2
Dong, A., Du, Z., & Yan, Z. (2019). Round trip time prediction using recurrent neural networks with minimal gated unit.
IEEE Communications Letters, 23(4), 584–587. doi:10.1109/LCOMM.2019.2899603
Dubovitskaya, A., Novotny, P., Xu, Z., & Wang, F. (2020). Applications of blockchain technology for data-sharing in
oncology: Results from a systematic literature review. Oncology, 98(6), 403–411. doi:10.1159/000504325 PMID:31794967
Dwivedi, R. K., Kumar, R., & Buyya, R. (2021). Secure healthcare monitoring sensor cloud with attribute-based el-
liptical curve cryptography. International Journal of Cloud Applications and Computing, 11(3), 1–18. doi:10.4018/
IJCAC.2021070101
Dwivedi, S., Vardhan, M., Tripathi, S., & Shukla, A. K. (2020). Implementation of adaptive scheme in evolutionary tech-
nique for anomaly-based intrusion detection. Evolutionary Intelligence, 13(1), 103–117. doi:10.100712065-019-00293-8
Edwards, C., Kouzani, A., Lee, K. H., & Ross, E. K. (n.d.). Neurostimulation Devices for the Treatment of Neurologic
Disorders. Mayo Clinic Proceedings, 92(9), 1427-1444. doi:10.1016/j.mayocp.2017.05.005
Edwards, S., & Profetis, I. (2016). Hajime: Analysis of a Decentralized Internet Worm for IoT Devices. http://security.
rapiditynetworks.com/publications/2016-10-16/Hajime.pdf
Ehdaie, M., Alexiou, N., Ahmadian, M., Aref, M. R., & Papadimitratos, P. (2017). Mitigating Node Capture Attack
in Random Key Distribution Schemes through Key Deletion. Journal of Communication Engineering, 6(2), 99–109.
Ehrenfeld, J. M. (2017). Wannacry, cybersecurity and health information technology: A time to act. Journal of Medical
Systems, 41(7), 104. doi:10.100710916-017-0752-1 PMID:28540616
Elhag, S., Fernández, A., Bawakid, A., Alshomrani, S., & Herrera, F. (2015). On the combination of genetic fuzzy systems
and pairwise learning for improving detection rates on intrusion detection systems. Expert Systems with Applications,
42(1), 193–202. doi:10.1016/j.eswa.2014.08.002
Elhoseny, M., Shankar, K., Lakshmanaprabu, S. K., Maseleno, A., & Arunkumar, N. (2020). Hybrid optimization with
cryptography encryption for medical image security in Internet of Things. Neural Computing & Applications, 32(15),
10979–10993.
Elnour, M., Meskin, N., Khan, K., & Jain, R. (2020). A dual-isolation-forests-based attack detection framework for
industrial control systems. IEEE Access: Practical Innovations, Open Solutions, 8, 36639–36651. doi:10.1109/AC-
CESS.2020.2975066
Eltayieb, N., Elhabob, R., Hassan, A., & Li, F. (2020). A blockchain-based attribute-based signcryption scheme to secure
data sharing in the cloud. Journal of Systems Architecture, 102, 101653. doi:10.1016/j.sysarc.2019.101653
Fabio Ceschini, G., Gatta, N., Venturini, M., Hubauer, T., & Murarasu, A. (2018). Optimization of statistical method-
ologies for anomaly detection in gas turbine dynamic time series. Journal of Engineering for Gas Turbines and Power,
140(3), 032401. doi:10.1115/1.4037963
270
Compilation of References
Fang, Z., Wang, J., Geng, J., & Kan, X. (2019). Feature Selection for Malware Detection Based on Reinforcement Learn-
ing. IEEE Access: Practical Innovations, Open Solutions, 7, 176177–176187. doi:10.1109/ACCESS.2019.2957429
Farnaaz, N., & Jabbar, M. A. (2016). Random forest modeling for network intrusion detection system. Procedia Computer
Science, 89, 213–217. doi:10.1016/j.procs.2016.06.047
Farooq, M., Waseem, M., Mazhar, S., Khairi, A., & Kamal, T. (2015). A Review on Internet of Things (IoT). International
Journal of Computers and Applications, 113(1), 1–7.
Feizollah, A., Anuar, N. B., Salleh, R., Suarez-Tangil, G., & Furnell, S. (2017). Androdialysis: Analysis of android intent
effectiveness in malware detection. Computers & Security, 65, 121-134.
Feizollah, A. (2014). Comparative study of k-means and mini batch k-means clustering algorithms in android malware
detection using network traffic analysis. In 2014 international symposium on biometrics and security technologies
(ISBAST). IEEE.
Feng, Q., He, D., Zeadally, S., Khan, M. K., & Kumar, N. (2019). A survey on privacy protection in blockchain system.
Journal of Network and Computer Applications, 126, 45–58. doi:10.1016/j.jnca.2018.10.020
Ficco, M., Esposito, C., Xiang, Y., & Palmieri, F. (2017). Pseudo-dynamic testing of realistic edge-fog cloud ecosystems.
IEEE Communications Magazine, 55(11), 98–104. doi:10.1109/MCOM.2017.1700328
Finke, A., Lenhardt, A., & Ritter, H. (2009). The MindGame: A P300-based brain–computer interface game. Neural
Networks, 22(9), 1329–1333. doi:10.1016/j.neunet.2009.07.003 PMID:19635654
Firdausi, I., Erwin, A., & Nugroho, A. S. (2010, December). Analysis of machine learning techniques used in behavior-
based malware detection. In 2010 second international conference on advances in computing, control, and telecom-
munication technologies (pp. 201-203). IEEE. 10.1109/ACT.2010.33
Fischer, T., & Krauss, C. (2018). Deep learning with long short-term memory networks for financial market predictions.
European Journal of Operational Research, 270(2), 654–669. doi:10.1016/j.ejor.2017.11.054
Florea, I., Ruse, L. C., & Rughinis, R. (2017, September). Challenges in security in Internet of Things. In 2017 16th
RoEduNet Conference: Networking in Education and Research (RoEduNet) (pp. 1-5). IEEE.
Frank, C., Nance, C., Jarocki, S., & Pauli, W. E. (2018). Protecting IoT from Mirai botnets; IoT device hardening. Journal
of Information Systems Applied Research, 11(2), 33–44.
Frank, M., Hwu, T., Jain, S., Knight, R. T., Martinovic, I., Mittal, P., Perito, D., Sluganovic, I., & Song, D. (2017) Us-
ing EEG-Based BCI Devices to Subliminally Probe for Private Information. Proceedings of the 2017 on Workshop on
Privacy in the Electronic Society - WPES ’17, 133-136. 10.1145/3139550.3139559
Fu, H.-L., Fang, P.-H., Chi, C.-Y., Kuo, C., Liu, M.-H., Hsu, H. M., Hsieh, C.-H., Liang, S.-F., Hsieh, S., & Yang, C.-T.
(2020). Application of Brain-Computer Interface and Virtual Reality in Advancing Cultural Experience. 2020 IEEE Interna-
tional Conference on Visual Communications and Image Processing (VCIP), 351–354. 10.1109/VCIP49819.2020.9301801
Gai, K., Wu, Y., Zhu, L., Qiu, M., & Shen, M. (2019). Privacy-preserving energy trading using consortium blockchain
in smart grid. IEEE Transactions on Industrial Informatics, 15(6), 3548–3558. doi:10.1109/TII.2019.2893433
Ganesh, M. (2017). CNN-based android malware detection. In 2017 International Conference on Software Security and
Assurance (ICSSA). IEEE. 10.1109/ICSSA.2017.18
Gangadharan, V., & Gupta, D. (2020). Recognizing Named Entities in Agriculture Documents using LDA based Topic
Modelling Techniques. Procedia Computer Science, 171, 1337–1345. doi:10.1016/j.procs.2020.04.143
271
Compilation of References
Garcia, S., Grill, M., Stiborek, J., & Zunino, A. (2014). An empirical comparison of botnet detection methods. Comput-
ers & Security, 45, 100-123.
Gaurav, A., Gupta, B. B., Hsu, C. H., Yamaguchi, S., & Chui, K. T. (2021, January). Fog Layer-based DDoS attack
Detection Approach for Internet-of-Things (IoTs) devices. In 2021 IEEE International Conference on Consumer Elec-
tronics (ICCE) (pp. 1-5). IEEE.
Gavriluţ, D., Cimpoeşu, M., Anton, D., & Ciortuz, L. (2009). Malware detection using machine learning. 2009 Inter-
national Multiconference on Computer Science and Information Technology, 735-741. 10.1109/IMCSIT.2009.5352759
Gelenbe, E., Domanska, J., Czàchorski, T., Drosou, A., & Tzovaras, D. (2018, June). Security for internet of things: The
seriot project. In 2018 International Symposium on Networks, Computers and Communications (ISNCC) (pp. 1-5). IEEE.
Ge, M., Hong, J. B., Guttmann, W., & Kim, D. S. (2017). A framework for automating security analysis of the internet
of things. Journal of Network and Computer Applications, 83, 12–27.
Ghoneim, A., Muhammad, G., Amin, S. U., & Gupta, B. (2018). Medical image forgery detection for smart healthcare.
IEEE Communications Magazine, 56(4), 33–37.
Gibert, D., Mateu, C., & Planes, J. (2019, July). A hierarchical convolutional neural network for malware classification.
In 2019 International Joint Conference on Neural Networks (IJCNN) (pp. 1-8). IEEE.
Gibert, D., Mateu, C., & Planes, J. (2020). The rise of machine learning for detection and classification of malware: Re-
search developments, trends and challenges. Journal of Network and Computer Applications, 153, 102526. doi:10.1016/j.
jnca.2019.102526
Gibert, D., Mateu, C., Planes, J., & Vicens, R. (2019). Using convolutional neural networks for classification of malware
represented as images. Journal of Computer Virology and Hacking Techniques, 15(1), 15–28.
Gilja, V., Nuyujukian, P., Chestek, C., Cunningham, J. P., Yu, B. M., Fan, J. M., Churchland, M. M., Kaufman, M. T.,
Kao, J. C., Ryu, S. I., & Shenoy, K. V. (2012). A high-performance neural prosthesis enabled by control algorithm design.
Nature Neuroscience, 15(12), 1752–1757. doi:10.1038/nn.3265 PMID:23160043
Glannon, G. (2014). Ethical issues with brain-computer interfaces. Frontiers in Systems Neuroscience, 8, 136. doi:10.3389/
fnsys.2014.00136 PMID:25126061
Gomes-Osman, J., Indahlastari, A., Fried, P. J., Cabral, D. L. F., Rice, J., Nissim, N. R., Aksu, S., McLaren, M. E., &
Woods, A. J. (2018). Non-invasive Brain Stimulation: Probing Intracortical Circuits and Improving Cognition in the
Aging Brain. Frontiers in Aging Neuroscience, 10, 177. doi:10.3389/fnagi.2018.00177 PMID:29950986
Gong, J., Xu, X., & Lei, Y. (2020). Unsupervised specific emitter identification method using radio-frequency finger-
print embedded InfoGAN. IEEE Transactions on Information Forensics and Security, 15, 2898–2913. doi:10.1109/
TIFS.2020.2978620
Gopal, T. S., Meerolla, M., Jyostna, G., Eswari, L., Reddy, P., & Magesh, E. (2018). Mitigating Mirai Malware Spreading
in IoT Environment. In Proc. of ICACCI 2018 (pp.2226-2230). Academic Press.
Gou, Z., & Yamaguchi, S. (2017). Analysis of various security issues and challenges in cloud computing environment:
a survey. In Identity Theft: Breakthroughs in Research and Practice (pp. 221-247). IGI Global.
Gou, Z., Yamaguchi, S., & Gupta, B. B. (2017). Analysis of various security issues and challenges in cloud computing
environment: a survey. In Identity Theft: Breakthroughs in Research and Practice (pp. 221-247). IGI Global.
272
Compilation of References
Gou, Z., Yamaguchi, S., & Gupta, B. B. (2017). Analysis of various security issues and challenges in cloud com-
puting environment: a survey. In Identity Theft: Breakthroughs in Research and Practice (pp. 221-247). IGI global.
doi:10.4018/978-1-5225-0808-3.ch011
Goyal, S. (2018). The History of Blockchain Technology: Must Know Timeline. Academic Press.
Grammatikis, P. I. R., Sarigiannidis, P. G., & Moscholios, I. D. (2019). Securing the Internet of Things: Challenges,
threats and solutions. Internet of Things, 5, 41–70.
Granjal, J., Monteiro, E., & Silva, J. S. (2015). Security for the internet of things: A survey of existing protocols and
open research issues. IEEE Communications Surveys and Tutorials, 17(3), 1294–1312.
Gruber, N., & Jockisch, A. (2020). Are GRU cells more specific and LSTM cells more sensitive in motive classification
of text? Frontiers in Artificial Intelligence, 3, 40. doi:10.3389/frai.2020.00040 PMID:33733157
Guan, C., Thulasidas, M., & Wu, J. (2004). High performance P300 speller for brain-computer interface. IEEE Interna-
tional Workshop on Biomedical Circuits and Systems (pp. S3/5/INV-S3/13). 10.1109/BIOCAS.2004.1454155
Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of things (IoT): A vision, architectural elements,
and future directions. Future Generation Computer Systems, 29(7), 1645–1660.
Gulzar, M., & Abbas, G. (2019, February). Internet of things security: a survey and taxonomy. In 2019 International
Conference on Engineering and Emerging Technologies (ICEET) (pp. 1-6). IEEE.
Guo, C., Huang, D., Dong, N., Zhang, J., & Xu, J. (2021). Callback2Vec: Callback-aware hierarchical embedding for
mobile application. Information Sciences, 542, 131–155. doi:10.1016/j.ins.2020.06.058
Guo, R., Shi, H., Zhao, Q., & Zheng, D. (2018). Secure attribute-based signature scheme with multiple authorities for
blockchain in electronic health records systems. IEEE Access: Practical Innovations, Open Solutions, 6, 11676–11686.
doi:10.1109/ACCESS.2018.2801266
Gupta, B. B., Li, K. C., Leung, V. C., Psannis, K. E., & Yamaguchi, S. (2021). Blockchain-assisted secure fine-grained
searchable encryption for a cloud-based healthcare cyber-physical system. IEEE/CAA Journal of Automatica Sinica.
Gupta, B. B., Gupta, S., Gangwar, S., Kumar, M., & Meena, P. K. (2015). Cross-site scripting (XSS) abuse and defense:
Exploitation on several testing bed environments and its defense. Journal of Information Privacy and Security, 11(2),
118–136.
Gupta, B. B., Perez, G. M., Agrawal, D. P., & Gupta, D. (2020). Handbook of computer networks and cyber security.
Springer. doi:10.1007/978-3-030-22277-2
Gupta, B. B., & Quamara, M. (2020). An overview of Internet of Things (IoT): Architectural aspects, challenges, and
protocols. Concurrency and Computation, 32(21), e4946.
Gupta, B. B., & Sheng, Q. Z. (Eds.). (2019). Machine learning for computer and cyber security: principle, algorithms,
and practices. CRC Press. doi:10.1201/9780429504044
Gupta, B. B., Yadav, K., Razzak, I., Psannis, K., Castiglione, A., & Chang, X. (2021). A novel approach for phishing
URLs detection using lexical based machine learning in a real-time environment. Computer Communications, 175,
47–57. doi:10.1016/j.comcom.2021.04.023
Gupta, B., Agrawal, D. P., & Yamaguchi, S. (2016). Handbook of Research on Modern Cryptographic Solutions for
Computer and Cyber Security. IGI Global. doi:10.4018/978-1-5225-0105-3
273
Compilation of References
Gupta, C. N., Palaniappan, R., & Paramesran, R. (2012). Exploiting the P300 paradigm for cognitive biometrics. Inter-
national Journal of Cognitive Biometrics, 1(1), 26. doi:10.1504/IJCB.2012.046513
Gupta, D., Gupta, B., & Mishra, A. (2018). Identity Theft, Malware, and Social Engineering in Dealing with Cybercrime.
In B. B. G. Gupta (Ed.), Computer and Cyber Security. CRC Press.
Gupta, K., & Shukla, S. (2016, February). Internet of Things: Security challenges for next generation networks. In 2016
International Conference on Innovation and Challenges in Cyber Security (ICICCS-INBUSH) (pp. 315-318). IEEE.
Gupta, S., & Gupta, B. B. (2015, May). PHP-sensor: a prototype method to discover workflow violation and XSS vulner-
abilities in PHP web applications. In Proceedings of the 12th ACM International Conference on Computing Frontiers
(pp. 1-8). ACM.
Gupta, S., & Gupta, B. B. (2016). JS‐SAN: Defense mechanism for HTML5‐based web applications against javascript
code injection vulnerabilities. Security and Communication Networks, 9(11), 1477–1495.
Gupta, S., & Gupta, B. B. (2016). XSS-SAFE: A server-side approach to detect and mitigate cross-site scripting (XSS)
attacks in JavaScript code. Arabian Journal for Science and Engineering, 41(3), 897–920.
Gupta, S., & Gupta, B. B. (2017). Detection, avoidance, and attack pattern mechanisms in modern web application
vulnerabilities: Present and future challenges. International Journal of Cloud Applications and Computing, 7(3), 1–43.
doi:10.4018/IJCAC.2017070101
HaddadPajouh, H., Dehghantanha, A., Khayami, R., & Choo, K.-K. R. (2018). A deep recurrent neural network based
approach for internet of things malware threat hunting. Future Generation Computer Systems, 85, 88–96. doi:10.1016/j.
future.2018.03.007
Hallappanavar, V. L., & Birje, M. N. (2021). A reliable trust computing mechanism in fog computing. International
Journal of Cloud Applications and Computing, 11(1), 1–20. doi:10.4018/IJCAC.2021010101
Hameed, S., Khan, F. I., & Hameed, B. (2019). Understanding security requirements and challenges in Internet of Things
(IoT): A review. Journal of Computer Networks and Communications.
Hao, X., Wang, Z., Shan, Z., & Zhao, Y. (2019). Prediction of electricity consumption in cement production: A
time-varying delay deep belief network prediction method. Neural Computing & Applications, 31(11), 7165–7179.
doi:10.100700521-018-3540-z
Harbi, Y., Aliouat, Z., Harous, S., Bentaleb, A., & Refoufi, A. (2019). A review of security in internet of things. Wireless
Personal Communications, 108(1), 325–344.
Hasan, M., Islam, M. M., Zarif, M. I. I., & Hashem, M. M. A. (2019). Attack and anomaly detection in IoT sensors in
IoT sites using machine learning approaches. Internet of Things, 7, 100059. doi:10.1016/j.iot.2019.100059
Hassan, M. M., Gumaei, A., Alsanad, A., Alrubaian, M., & Fortino, G. (2020). A hybrid deep learning model for ef-
ficient intrusion detection in big data environment. Information Sciences, 513, 386–396. doi:10.1016/j.ins.2019.10.069
Hassan, W. H. (2019). Current research on Internet of Things (IoT) security: A survey. Computer Networks, 148, 283–294.
doi:10.1016/j.comnet.2018.11.025
Hasselgren, A., Kralevska, K., Gligoroski, D., Pedersen, S. A., & Faxvaag, A. (2019). Blockchain in healthcare and health
sciences–a scoping review. International Journal of Medical Informatics, 104040. PMID:31865055
274
Compilation of References
Hastie, T., Tibshirani, R., & Friedman, J. (2009). The elements of statistical learning: data mining, inference, and predic-
tion. Springer Science & Business Media. Industrial Control System (ICS) Cyber Attack Datasets. https://sites.google.
com/a/uah.edu/tommy-morris-uah/ics-data-sets
Hiraishi, K. (2001). A Petri-net-based model for the mathematical analysis of multi-agent systems. IEICE Trans. on
Fundamentals, E84-A(11), 2829–2837.
Hofman, W. J. (2019). A Methodological Approach for Development and Deployment of Data Sharing in Complex
Organizational Supply and Logistics Networks with Blockchain Technology. IFAC-PapersOnLine, 52(3), 55–60.
doi:10.1016/j.ifacol.2019.06.010
Hong, J., & Liu, C. C. (2017). Intelligent electronic devices with collaborative intrusion detection systems. IEEE Trans-
actions on Smart Grid, 10(1), 271–281. doi:10.1109/TSG.2017.2737826
Hong, K., & Khan, M. J. (2017). Hybrid Brain–Computer Interface Techniques for Improved Classification Accuracy
and Increased Number of Commands: A Review. Frontiers in Neurorobotics, 11, 35. doi:10.3389/fnbot.2017.00035
PMID:28790910
Hossain, K., Rahman, M., & Roy, S. (2019). Iot data compression and optimization techniques in cloud storage: Current
prospects and future directions. International Journal of Cloud Applications and Computing, 9(2), 43–59.
Hossain, M. M., Fotouhi, M., & Hasan, R. (2015, June). Towards an analysis of security issues, challenges, and open
problems in the internet of things. In 2015 IEEE world congress on services. IEEE.
Hou, J., Qu, L., & Shi, W. (2019). A survey on internet of things security from data perspectives. Computer Networks,
148, 295–306.
Hu, C., Zhang, J., & Wen, Q. (2011, October). An identity-based personal location system with protected privacy in IoT.
In 2011 4th IEEE International Conference on Broadband Network and Multimedia Technology (pp. 192-195). IEEE.
Huang, W., & Stokes, J. W. (2016, July). MtNet: a multi-task neural network for dynamic malware classification. In
International conference on detection of intrusions and malware, and vulnerability assessment (pp. 399-418). Springer.
10.1007/978-3-319-40667-1_20
Huertas Celdrán, A., Gil Pérez, M., Mlakar, I., Alcaraz Calero, J. M., García Clemente, F. J., & Martínez Pérez, G.
(2019). A Management Platform for Citizen’s Data Protection Regulation. In G. Wang, A. El Saddik, X. Lai, G. Martinez
Perez, & K. K. Choo (Eds.), Smart City and Informatization. iSCI 2019. Communications in Computer and Information
Science (Vol. 1122). Springer. doi:10.1007/978-981-15-1301-5_6
Huertas Celdrán, A., Gil Pérez, M., Mlakar, I., Alcaraz Calero, J. M., García Clemente, F. J., Martínez Pérez, G., & Bhui-
yan, Z. A. (2020). PROTECTOR: Towards the protection of sensitive data in Europe and the US. Computer Networks,
181, 107448. doi:10.1016/j.comnet.2020.107448
Husamuddin, M., & Qayyum, M. (2017, March). Internet of Things: A study on security and privacy threats. In 2017
2nd International Conference on Anti-Cyber Crimes (ICACC) (pp. 93-97). IEEE.
Idika, N., & Mathur, A. P. (2007). A survey of malware detection techniques. Purdue University.
Ienca, M., & Haselager, P. (2016). Hacking the brain: Brain–computer interfacing technology and the ethics of neuros-
ecurity. Ethics and Information Technology, 18(2), 117–129. doi:10.100710676-016-9398-9
Ienca, M., Haselager, P., & Emanuel, E. (2018). Brain leaks and consumer neurotechnology. Nature Biotechnology,
36(9), 805–810. doi:10.1038/nbt.4240 PMID:30188521
275
Compilation of References
Imad, J., Mohammed, F., Jaroodi, J. A., & Mohamed, N. (2016). TRAS: a trust-based routing protocol for ad hoc and
sensor networks. IEEE 2nd international conference on big data security on cloud, IEEE international conference on
high performance and smart computing, IEEE international conference on intelligent data and security, 382–387.
Inoue, J., Yamagata, Y., Chen, Y., Poskitt, C. M., & Sun, J. (2017, November). Anomaly detection for a water treat-
ment system using unsupervised machine learning. In 2017 IEEE International Conference on Data Mining Workshops
(ICDMW) (pp. 1058-1065). IEEE. http://kdd.ics.uci.edu/databases/kddcup99/kddcup99.html
Iscan, H., & Gunduz, M. (2015, November). A survey on fruit fly optimization algorithm. In 2015 11th International
Conference on Signal-Image Technology & Internet-Based Systems (SITIS) (pp. 520-527). 10.1109/SITIS.2015.55
Iturrate, I., Antelis, J. M., Kubler, A., & Minguez, J. (2009). A Non-invasive Brain-Actuated Wheelchair Based on a P300
Neurophysiological Protocol and Automated Navigation. IEEE Transactions on Robotics, 25(3), 614–627. doi:10.1109/
TRO.2009.2020347
Jamali, S., & Jafarzadeh, P. (2017). An intelligent intrusion detection system by using hierarchically structured learning
automata. Neural Computing & Applications, 28(5), 1001–1008. doi:10.100700521-015-2116-4
Jang, Y. S., Ryu, S. A., & Park, K. C. (2011). Analysis of P300 Related Target Choice in Oddball Paradigm. Journal of
Information and Communication Convergence Engineering, 9(2), 125–128. doi:10.6109/jicce.2011.9.2.125
Jaramillo, L. E. S. (2018). Malware Detection and Mitigation Techniques: Lessons Learned from Mirai DDOS Attack.
Journal of Information Systems Engineering & Management, 3(3), 19.
Jerbi, W., Guermazi, A., & Trabelsi, H. (2020). A novel secure routing protocol of generation and management cryp-
tographic keys for wireless sensor networks deployed in internet of things. International Journal of High Performance
Computing and Networking, 16(2-3), 87–94.
Jerlin, M. A., & Marimuthu, K. (2018). A new malware detection system using machine learning techniques for API call
sequences. Journal of Applied Security Research, 13(1), 45–62. doi:10.1080/19361610.2018.1387734
Jindal, F., Jamar, R., & Churi, P. (2018). Future and challenges of internet of things. International Journal of Computer
Science & Information Technology, 10(2), 13–25.
Jose, D. V., & Vijyalakshmi, A. (2018). An overview of security in Internet of Things. Procedia Computer Science, 143,
744–748. doi:10.1016/j.procs.2018.10.439
Kalyani, G., & Chaudhari, S. (2020). An efficient approach for enhancing security in Internet of Things using the opti-
mum authentication key. International Journal of Computers and Applications, 42(3), 306–314.
Kancherla, K., & Mukkamala, S. (2013, April). Image visualization based malware detection. In 2013 IEEE Symposium
on Computational Intelligence in Cyber Security (CICS) (pp. 40-44). IEEE.
Kaongoen, N., Yu, M., & Jo, S. (2020). Two-Factor Authentication System Using P300 Response to a Sequence of
Human Photographs. IEEE Transactions on Systems, Man, and Cybernetics. Systems, 50(3), 1178–1185. doi:10.1109/
TSMC.2017.2756673
Kaspersky. (2021). Machine Learning for Malware Detection. Retrieved from https://media.kaspersky.com/en/enterprise-
security/Kaspersky-Lab-Whitepaper-Machine-Learning.pdf
Kaur, J., Gill, S. S., & Dhaliwal, B. S. (2016). Secure trust based key management routing framework for wireless sensor
networks. Journal of Engineering (Stevenage, England), 2016, 1–9.
276
Compilation of References
Keyes, L., & Kaur, L. Gagnon, & Massicotte. (2021). EntropLyzer: Android Malware Classification and Characterization
Using Entropy Analysis of Dynamic Characteristics. In Reconciling Data Analytics, Automation, Privacy, and Security:
A Big Data Challenge (RDAAPS). IEEE.
Khabarova, E. A., Denisova, N. P., Dmitriev, A. B., Slavin, K. V., & Verhagen Metman, L. (2018). Deep Brain, Stimu-
lation of the Subthalamic Nucleus in Patients with Parkinson Disease with Prior Pallidotomy or Thalamotomy. Brain
Sciences, 8(4), 66. doi:10.3390/brainsci8040066 PMID:29659494
Khalid, A., Kirisci, P., Ghrairi, Z., Thoben, K. D., & Pannek, J. (2017, July). Towards implementing safety and security
concepts for human-robot collaboration in the context of Industry 4.0. In 39th International MATADOR Conference on
Advanced Manufacturing (Manchester, UK).
Khalid, A., Kirisci, P., Khan, Z. H., Ghrairi, Z., Thoben, K. D., & Pannek, J. (2018). Security framework for industrial
collaborative robotic cyber-physical systems. Computers in Industry, 97, 132–145. doi:10.1016/j.compind.2018.02.009
Khan, M., Wang, H., & Ngueilbaye, A. (2021). Attention-Based Deep Gated Fully Convolutional End-to-End Architec-
tures for Time Series Classification. Neural Processing Letters, 1–34.
Khan, Z. H., Khalid, A., & Iqbal, J. (2018). Towards realizing robotic potential in future intelligent food manufacturing
systems. Innovative Food Science & Emerging Technologies, 48, 11–24. doi:10.1016/j.ifset.2018.05.011
Khattak, H. A., Shah, M. A., Khan, S., Ali, I., & Imran, M. (2019). Perception layer security in Internet of Things. Future
Generation Computer Systems, 100, 144–164. doi:10.1016/j.future.2019.04.038
Kim. (2018). A multimodal deep learning method for android malware detection using various features. IEEE Transac-
tions on Information Forensics and Security, 14(3), 773-788.
Kolias, C., Kambourakis, G., Stavrou, A., & Voas, J. (2017). DDoS in the IoT: Mirai and other botnets. IEEE Computer,
50(7), 80–84. doi:10.1109/MC.2017.201
Kolias, C., Stavrou, A., Voas, J., Bojanova, I., & Kuhn, R. (2016). Learning Internet-of-Things security” hands-on. IEEE
Security and Privacy, 14(1), 37–46.
Kolosnjaji, B., Zarras, A., Webster, G., & Eckert, C. (2016, December). Deep learning for classification of malware
system call sequences. In Australasian Joint Conference on Artificial Intelligence (pp. 137-149). Springer. 10.1007/978-
3-319-50127-7_11
Kosba, A., Miller, A., Shi, E., Wen, Z., & Papamanthou, C. (2016, May). Hawk: The blockchain model of cryptography
and privacy-preserving smart contracts. In 2016 IEEE symposium on security and privacy (SP) (pp. 839-858). IEEE.
Kouicem, D. E., Bouabdallah, A., & Lakhlef, H. (2018). Internet of things security: A top-down survey. Computer
Networks, 141, 199–221.
Kravchik, M., & Shabtai, A. (2018, January). Detecting cyber attacks in industrial control systems using convolutional
neural networks. In Proceedings of the 2018 Workshop on Cyber-Physical Systems Security and PrivaCy (pp. 72-83).
10.1145/3264888.3264896
Kravchik, M., & Shabtai, A. (2021). Efficient cyber attack detection in industrial control systems using lightweight neural
networks and PCA. IEEE Transactions on Dependable and Secure Computing, 1. doi:10.1109/TDSC.2021.3050101
Krčál, M., Švec, O., Bálek, M., & Jašek, O. (2018). Deep convolutional malware classifiers can learn from raw ex-
ecutables and labels only. Academic Press.
277
Compilation of References
Krichen, M., Cheikhrouhou, O., Lahami, M., Alroobaea, R., & Maâlej, A. J. (2017, November). Towards a model-based
testing framework for the security of internet of things for smart city applications. In International Conference on Smart
Cities, Infrastructure, Technologies and Applications (pp. 360-365). Springer.
Kshetri, N., & Voas, J. (2017). Hacking power grids: A current problem. Computer, 50(12), 91–95. doi:10.1109/
MC.2017.4451203
Kubanek, J. (2018). Neuromodulation with transcranial focused ultrasound. Neurosurgical Focus FOC, 44(2), E14.
doi:10.3171/2017.11.FOCUS17621 PMID:29385924
Kumar, P., Gupta, G. P., & Tripathi, R. (2021). Toward Design of an Intelligent Cyber Attack Detection System using
Hybrid Feature Reduced Approach for IoT Networks. Arabian Journal for Science and Engineering, 46(4), 3749–3778.
doi:10.100713369-020-05181-3
Kumar, S. (2020). An emerging threat Fileless malware: A survey and research challenges. Cybersecurity, 3(1), 1–12.
doi:10.118642400-019-0043-x
Landau, O., Puzis, R., & Nissum, N. (2020). Mind Your Mind: EEG-Based Brain-Computer Interfaces and Their Security
in Cyber Space. ACM Computing Surveys, 53(1), 1–38. doi:10.1145/3372043
Langner, R. (2011). Stuxnet: Dissecting a cyberwarfare weapon. IEEE Security and Privacy, 9(3), 49–51. doi:10.1109/
MSP.2011.67
Larochelle, H., Mandel, M., Pascanu, R., & Bengio, Y. (2012). Learning algorithms for the classification restricted
boltzmann machine. Journal of Machine Learning Research, 13(1), 643–669.
Lashkari, A. H. (2018). Toward developing a systematic approach to generate benchmark android malware datasets
and classification. In 2018 International Carnahan Conference on Security Technology (ICCST). IEEE. 10.1109/
CCST.2018.8585560
Lasi, H., Fettke, P., Kemper, H. G., Feld, T., & Hoffmann, M. (2014). Industry 4.0. Business & Information Systems
Engineering, 6(4), 239–242. doi:10.100712599-014-0334-4
Lavin, A., & Ahmad, S. (2015, December). Evaluating Real-Time anomaly detection algorithms--The Numenta anomaly
benchmark. In 2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA) (pp. 38-44).
IEEE.
Learning-Based Security Technique for Selective Forwarding Attack in Clustered WSN. (n.d.). Applied Soft Computing,
108, 107473. doi:10.1016/j.asoc.2021.107473
Lebedev, M. A., & Nicolelis, A. L. (2017). Brain-Machine Interfaces: From Basic Science to Neuroprostheses and Neu-
rorehabilitation. Physiological Reviews, 97(2), 767–837. doi:10.1152/physrev.00027.2016 PMID:28275048
Lee, J., Mok, E., Huang, J., Cui, L., Lee, A., Leung, V., Mercier, P., Shellhammer, S., Larson, L., Asbeck, P., Rao, R.,
Song, Y., Nurmikko, A., & Laiwalla, F. (2019) An Implantable Wireless Network of Distributed Microscale Sensors
for Neural Applications. 2019 9th International IEEE/EMBS Conference on Neural Engineering (NER) (pp. 871-874).
10.1109/NER.2019.8717023
Lee, J. H., & Kim, H. (2017). Security and privacy challenges in the internet of things [security and privacy matters].
IEEE Consumer Electronics Magazine, 6(3), 134–136.
Lee, J., Park, D., & Lee, C. (2017). Feature selection algorithm for intrusions detection system using sequential forward
search and random forest classifier. Transactions on Internet and Information Systems (Seoul), 11(10), 5132–5148.
278
Compilation of References
Legon, W., Bansal, P., Tyshynsky, R., Ai, L., & Mueller, J. K. (2018). Transcranial focused ultrasound neuromodulation
of the human primary motor cortex. Scientific Reports, 8(1), 10007. doi:10.103841598-018-28320-1 PMID:29968768
Lei, Y., Yang, B., Jiang, X., Jia, F., Li, N., & Nandi, A. K. (2020). Applications of machine learning to machine fault diag-
nosis: A review and roadmap. Mechanical Systems and Signal Processing, 138, 106587. doi:10.1016/j.ymssp.2019.106587
Letteri, I., Penna, G. D., & Gasperis, G. D. (2019). Security in the internet of things: Botnet detection in software-defined
networks by deep learning techniques. International Journal of High Performance Computing and Networking, 15(3-4),
170–182.
Liang, W., Tang, M., Long, J., Peng, X., Xu, J., & Li, K. C. (2019). A secure fabric blockchain-based data transmission
technique for industrial Internet-of-Things. IEEE Transactions on Industrial Informatics, 15(6), 3582–3592. doi:10.1109/
TII.2019.2907092
Liao, H. J., Lin, C. H. R., Lin, Y. C., & Tung, K. Y. (2013). Intrusion detection system: A comprehensive review. Journal
of Network and Computer Applications, 36(1), 16–24. doi:10.1016/j.jnca.2012.09.004
Liao, K., Lin, C., Zhao, Y., & Gabbouj, M. (2019). DR-GAN: Automatic radial distortion rectification using conditional
GAN in real-time. IEEE Transactions on Circuits and Systems for Video Technology, 30(3), 725–733. doi:10.1109/
TCSVT.2019.2897984
Liberati, G., da Rocha, J. L. D., van der Heiden, L., Raffone, A., Birbaumer, N., Olivetti Belardinelli, M., & Sitaram,
R. (2012). Toward a Brain-Computer Interface for Alzheimer’s Disease Patients by Combining Classical Conditioning
and Brain State Classification. Journal of Alzheimer’s Disease, 31(s3), S211–S220. doi:10.3233/JAD-2012-112129
PMID:22451316
Li, D., Chen, D., Jin, B., Shi, L., Goh, J., & Ng, S. K. (2019, September). MAD-GAN: Multivariate anomaly detection
for time series data with generative adversarial networks. In International Conference on Artificial Neural Networks (pp.
703-716). Springer. 10.1007/978-3-030-30490-4_56
Lien, J., Amihood, P. M., Javidan, A. J., Karagozler, M. E., Olson, E. M., & Poupyrev, I. (2017). Embedding Radars in
Robots for Safety and Obstacle Detection. Academic Press.
Li, F., Shi, Y., Shinde, A., Ye, J., & Song, W. (2019). Enhanced cyber-physical security in internet of things through
energy auditing. IEEE Internet of Things Journal, 6(3), 5224–5231.
Li, F., & Xiong, P. (2013). Practical secure communication for integrating wireless sensor networks into the internet of
things. IEEE Sensors Journal, 13(10), 3677–3684.
Li, J., Sun, L., Yan, Q., Li, Z., Srisa-An, W., & Ye, H. (2018). Significant permission identification for machine-learn-
ing-based android malware detection. IEEE Transactions on Industrial Informatics, 14(7), 3216–3225. doi:10.1109/
TII.2017.2789219
Lin & Guowei Wu. (2013). Enhancing the attacking efficiency of the node captureattack in WSN: a matrix approach. J
Supercomput, Springer Science &Business Media, 1-19.
Lin, J., Yu, W., Zhang, N., Yang, X., Zhang, H., & Zhao, W. (2017). A survey on internet of things: Architecture, enabling
technologies, security and privacy, and applications. IEEE Internet of Things Journal, 4(5), 1125-1142.
Lin, C., Qiu, T., Obaidat, M. S., Yu, C. W., Yao, L., & Wu, G. (2016). MREA: A minimum resource expenditure node
capture attack in wireless sensor networks. Security and Communication Networks, 9(18), 5502–5517. doi:10.1002ec.1713
279
Compilation of References
Lin, C., Wu, G., Yu, C. W., & Yao, L. (2015). Maximizing destructiveness of node capture attack in wireless sensor
networks. The Journal of Supercomputing, 71(8), 3181–3212. doi:10.100711227-015-1435-7
Li, P., Salour, M., & Su, X. (2008). A survey of internet worm detection and containment. IEEE Communications Surveys
and Tutorials, 10(1), 20–35. doi:10.1109/COMST.2008.4483668
Li, Q., Ding, D., & Conti, M. (2015). Brain-Computer Interface applications: Security and privacy challenges. 2015
IEEE Conference on Communications and Network Security (CNS), 663-666. 10.1109/CNS.2015.7346884
Li, S., & Da Xu, L. (2017). Securing the internet of things. Syngress.
Li, S., Da Xu, L., & Zhao, S. (2015). The internet of things: A survey. Information Systems Frontiers, 17(2), 243–259.
Li, S., Tryfonas, T., & Li, H. (2016). The Internet of Things: A security point of view. Internet Research.
Liu, C. H., Lin, Q., & Wen, S. (2018). Blockchain-enabled data collection and sharing for industrial IoT with deep re-
inforcement learning. IEEE Transactions on Industrial Informatics, 15(6), 3516–3526. doi:10.1109/TII.2018.2890203
Liu, L., Wang, B., Yu, B., & Zhong, Q. (2017). Automatic malware classification and new malware detection using machine
learning. Frontiers of Information Technology & Electronic Engineering, 18(9), 1336–1347. doi:10.1631/FITEE.1601325
Liu, Q., Sun, L., Kornhauser, A., Sun, J., & Sangwa, N. (2019). Road roughness acquisition and classification using
improved restricted Boltzmann machine deep learning algorithm. Sensor Review, 39(6), 733–742. doi:10.1108/SR-05-
2018-0132
Liu, X., Wang, Z., Jin, C., Li, F., & Li, G. (2019). A Blockchain-Based Medical Data Sharing and Protection Scheme.
IEEE Access: Practical Innovations, Open Solutions, 7, 118943–118953. doi:10.1109/ACCESS.2019.2937685
Liu, X., Zhao, M., Li, S., Zhang, F., & Trappe, W. (2017). A security framework for the internet of things in the future
internet architecture. Future Internet, 9(3), 27.
Liu, Y., Kuang, Y., Xiao, Y., & Xu, G. (2017). SDN-based data transfer security for Internet of Things. IEEE Internet
of Things Journal, 5(1), 257–268. doi:10.1109/JIOT.2017.2779180
Liu, Y., Zhang, J., & Gao, Q. (2018, October). A Blockchain-Based Secure Cloud Files Sharing Scheme with Fine-
Grained Access Control. In 2018 International Conference on Networking and Network Applications (NaNA) (pp. 277-
283). IEEE. 10.1109/NANA.2018.8648778
Li, X., Hu, Z., Xu, M., Wang, Y., & Ma, J. (2021). Transfer learning based intrusion detection scheme for Internet of
vehicles. Information Sciences, 547, 119–135. doi:10.1016/j.ins.2020.05.130
Li, Z., Yin, X., Geng, Z., Zhang, H., Li, P., Sun, Y., ... Li, L. (2013, January). Research on PKI-like Protocol for the
Internet of Things. In 2013 Fifth International Conference on Measuring Technology and Mechatronics Automation
(pp. 915-918). IEEE.
López Bernal, S., Huertas Celdrán, A., & Martínez Pérez, G. (2019). Cybersecurity on Brain-Computer Interfaces: at-
tacks and countermeasures. V Jornadas Nacionales de Investigación en Ciberseguridad (JNIC2019), 198-199.
López Bernal, S., Huertas Celdrán, A., Fernández Maimó, L., Martínez Pérez, G., Barros, M. T., & Balasubramaniam,
S. (2020). Cyberattacks on Miniature Brain Implants to Disrupt Spontaneous Neural Signaling. IEEE Access: Practical
Innovations, Open Solutions, 8, 152204–152222. doi:10.1109/ACCESS.2020.3017394
López Bernal, S., Huertas Celdrán, A., Martínez Pérez, G., Barros, M. T., & Balasubramaniam, S. (2021). Security in
Brain-Computer Interfaces: State-of-the-Art, Opportunities, and Future Challenges. ACM Computing Surveys, 54(1),
35. doi:10.1145/3427376
280
Compilation of References
Lü, X., Meng, L., Chen, C., & Wang, P. (2019). Fuzzy removing redundancy restricted boltzmann machine: Improv-
ing learning speed and classification accuracy. IEEE Transactions on Fuzzy Systems, 28(10), 2495–2509. doi:10.1109/
TFUZZ.2019.2940415
Lu, Y., Huang, X., Dai, Y., Maharjan, S., & Zhang, Y. (2019). Blockchain and Federated Learning for Privacy-preserved
Data Sharing in Industrial IoT. IEEE Transactions on Industrial Informatics.
Lv, Z. (2020). Security of internet of things edge devices. Software, Practice & Experience, spe.2806. doi:10.1002pe.2806
Lyu, X., Ding, Y., & Yang, S. H. (2019). Safety and security risk assessment in cyber-physical systems. IET Cyber-
Physical Systems. Theory & Applications, 4(3), 221–232.
Madakam, S., Lake, V., Lake, V., & Lake, V. (2015). Internet of Things (IoT): A literature review. Journal of Computer
and Communications, 3(05), 164.
Mahdavifar, Fitriah Kadir, Fatemi, Alhadidi, & Ghorbani. (2020). Dynamic Android Malware Category Classification
using Semi-Supervised Deep Learning. The 18th IEEE International Conference on Dependable, Autonomic, and Secure
Computing (DASC).
Mahmoud, R., Yousuf, T., Aloul, F., & Zualkernan, I. (2015, December). Internet of things (IoT) security: Current sta-
tus, challenges and prospective measures. In 2015 10th International Conference for Internet Technology and Secured
Transactions (ICITST) (pp. 336-341). IEEE.
Makhdoom, I., Abolhasan, M., Lipman, J., Liu, R. P., & Ni, W. (2018). Anatomy of threats to the internet of things.
IEEE Communications Surveys and Tutorials, 21(2), 1636–1675.
Ma, L., Liu, Y., Zhang, X., Ye, Y., Yin, G., & Johnson, B. A. (2019). Deep learning in remote sensing applications:
A meta-analysis and review. ISPRS Journal of Photogrammetry and Remote Sensing, 152, 166–177. doi:10.1016/j.
isprsjprs.2019.04.015
Mani, N., Moh, M., & Moh, T. S. (2021). Defending deep learning models against adversarial attacks. International
Journal of Software Science and Computational Intelligence, 13(1), 72–89.
Manso, P., Moura, J., & Serrão, C. (2019). SDN-Based Intrusion Detection System for Early Detection and Mitigation
of DDoS Attacks. Information, 10, 106.
Manzoor, A., Liyanage, M., Braeke, A., Kanhere, S. S., & Ylianttila, M. (2019, May). Blockchain based proxy re-encryption
scheme for secure IoT data sharing. In 2019 IEEE International Conference on Blockchain and Cryptocurrency (ICBC)
(pp. 99-103). IEEE. 10.1109/BLOC.2019.8751336
Maple, C. (2017). Security and privacy in the internet of things. Journal of Cyber Policy, 2(2), 155–184.
Martinovic, I., Davies, D., & Frank, M. (2012). On the feasibility of side-channel attacks with brain-computer interfaces.
Proceedings of the 21st USENIX Security Symposium, 143-158.
Mastorakis, G., Mavromoustakis, C. X., & Pallis, E. (2017). Beyond the internet of things (J. M. Batalla, Ed.). Springer.
Masud, M., Gaba, G. S., Alqahtani, S., Muhammad, G., Gupta, B. B., Kumar, P., & Ghoneim, A. (2020). A lightweight
and robust secure key establishment protocol for internet of medical things in COVID-19 patients care. IEEE Internet
of Things Journal.
Mathur, A. P., & Tippenhauer, N. O. (2016, April). SWaT: a water treatment testbed for research and training on ICS se-
curity. In 2016 international workshop on cyber-physical systems for smart water networks (CySWater) (pp. 31-36). IEEE.
MATLAB - El lenguaje del cálculo técnico. (2020). MATLAB & Simulink. https://es.mathworks.com/products/matlab.html
281
Compilation of References
Mattern, F., & Floerkemeier, C. (2010). From the Internet of Computers to the Internet of Things. In From active data
management to event-based systems and more (pp. 242–259). Springer.
Mavropoulos, O., Mouratidis, H., Fish, A., & Panaousis, E. (2019). Apparatus: A framework for security analysis in
internet of things systems. Ad Hoc Networks, 92, 101743.
McEwen, A., & Cassimally, H. (2013). Designing the internet of things. John Wiley & Sons.
McGhin, T., Choo, K. K. R., Liu, C. Z., & He, D. (2019). Blockchain in healthcare applications: Research challenges
and opportunities. Journal of Network and Computer Applications, 135, 62–75. doi:10.1016/j.jnca.2019.02.027
McMahon, M., & Schukat, M. (2018). A low-Cost, Open-Source, BCI- VR Game Control Development Environment
Prototype for Game Based Neurorehabilitation. IEEE Games, Entertainment, Media Conference (GEM), 1-9. 10.1109/
GEM.2018.8516468
Meidan, Y., Bohadana, M., Mathov, Y., Mirsky, Y., Shabtai, A., Breitenbacher, D., & Elovici, Y. (2018). N-BaIoT -
Network-Based Detection of IoT Botnet Attacks Using Deep Autoencoders. IEEE Pervasive Computing, 17(3), 12–22.
Mendez Mena, D., Papapanagiotou, I., & Yang, B. (2018). Internet of things: Survey on security. Information Security
Journal: A Global Perspective, 27(3), 162-182.
Mendez, D. M., Papapanagiotou, I., & Yang, B. (2017). Internet of things: Survey on security and privacy. arXiv preprint
arXiv:1707.01879.
Mienye, I. D., Sun, Y., & Wang, Z. (2020). Improved sparse autoencoder based artificial neural network approach for
prediction of heart disease. Informatics in Medicine Unlocked, 18, 100307. doi:10.1016/j.imu.2020.100307
Milić, J. (2019). Mirai Botnet Continues to Plague IoT Space. Retrieved from https://blog.reversinglabs.com/blog/mirai-
botnet-continues-to-plague-iot-space
Miller, B., & Rowe, D. (2012, October). A survey SCADA of and critical infrastructure incidents. In Proceedings of the
1st Annual conference on Research in information technology (pp. 51-56). 10.1145/2380790.2380805
Miloslavskaya, N., & Tolstoy, A. (2019). Internet of things: Information security challenges and solutions. Cluster
Computing, 22(1), 103–119.
Mirian, A., Ma, Z., Adrian, D., Tischer, M., Chuenchujit, T., & Yardley, T. (2016, December). An internet-wide view
of ICS devices. In 2016 14th Annual Conference on Privacy, Security and Trust (PST) (pp. 96-103). IEEE. 10.1109/
PST.2016.7906943
Mirsadeghi, F., Rafsanjani, M. K., & Gupta, B. B. (2020). A trust infrastructure based authentication method for clustered
vehicular ad hoc networks. Peer-to-Peer Networking and Applications, 1–17.
Mishra, A., Gupta, N., & Gupta, B. B. (2021). Defense mechanisms against DDoS attack based on entropy in SDN-cloud
using POX controller. Telecommunication Systems, 1–16.
Misra, S., Maheswaran, M., & Hashmi, S. (2017). Security challenges and approaches in internet of things. Springer
International Publishing.
Moffitt, T. (2016). Source Code for Mirai IoT Malware Released. Retrieved from https://www.webroot.com/blog/2016/10/10/
source-code-Mirai-iot-malware-released/
Mohan, A. P., & Gladston, A. (2020). Merkle tree and Blockchain-based cloud data auditing. International Journal of
Cloud Applications and Computing, 10(3), 54–66. doi:10.4018/IJCAC.2020070103
282
Compilation of References
Mohanta, B. K., Jena, D., Panda, S. S., & Sobhanayak, S. (2019). Blockchain Technology: A Survey on Applications
and Security Privacy Challenges. Internet of Things, 100107.
Molesky, M. J., & Cameron, E. A. (2019). Internet of Things: An Analysis and Proposal of White Worm Technology.
In Proc. of IEEE ICCE 2019. Academic Press.
Moore, D., Paxson, V., Savage, S., Shannon, C., Staniford, S., & Weaver, N. (2003). The spread of the sapphire/slammer
worm. CAIDA, ICSI, Silicon Defense, UC Berkeley EECS and UC San Diego CSE.
Moreno, I., Batista, E., Serracin, S., Moreno, R., Gómez, L., Serracin, J., Quintero, J., & Boya, C. (2019). Los siste-
mas de interfaz cerebro-computadora basado en EEG: características y aplicaciones. I+D Tecnológico, 15(2), 13–26.
doi:10.33412/idt.v15.2.2230
Morris, T., & Gao, W. (2014, March). Industrial control system traffic data sets for intrusion detection research. In
International Conference on Critical Infrastructure Protection (pp. 65-78). Springer. 10.1007/978-3-662-45355-1_5
Morris, T., Srivastava, A., Reaves, B., Gao, W., Pavurapu, K., & Reddi, R. (2011). A control system testbed to validate
critical infrastructure protection concepts. International Journal of Critical Infrastructure Protection, 4(2), 88–103.
doi:10.1016/j.ijcip.2011.06.005
Movahedi, F., Coyle, J. L., & Sejdić, E. (2017). Deep belief networks for electroencephalography: A review of recent
contributions and future outlooks. IEEE Journal of Biomedical and Health Informatics, 22(3), 642–652. doi:10.1109/
JBHI.2017.2727218 PMID:28715343
Muna, A. H., Moustafa, N., & Sitnikova, E. (2018). Identification of malicious activities in industrial internet of things
based on deep learning models. Journal of Information Security and Applications, 41, 1-11.
Munro, K. (2012). Deconstructing flame: The limitations of traditional defences. Computer Fraud & Security, 2012(10),
8–11. doi:10.1016/S1361-3723(12)70102-1
Murata, T. (1989). Petri nets: Properties, analysis and applications. Proceedings of the IEEE, 77(4), 541–580.
Musk, E. (2019). Neuralink An Integrated Brain-Machine Interface Platform With Thousands of Channels. Journal of
Medical Internet Research, 21(10), e16194. doi:10.2196/16194 PMID:31642810
Na, E., Lee, K., Kim, E. J., Bae, J. B., Suh, S. W., Byun, S., Han, J. W., & Kim, K. W. (2021). Pre-attentive Visual Pro-
cessing in Alzheimer’s Disease: An Event-related Potential Study. Current Alzheimer Research, 17(13), 1195–1207. do
i:10.2174/1567205018666210216084534 PMID:33593259
Nakahori, K., & Yamaguchi, S. (2017). A support tool to design IoT services with NuSMV. In Proc. of IEEE ICCE
2017 (pp.84–87). IEEE.
Nakao, K. (2018). Proactive cyber security response by utilizing passive monitoring technologies. In Proc. of IEEE ICCE
2018 (p. 1). 10.1109/ICCE.2018.8326061
Nakasumi, M. (2017, July). Information sharing for supply chain management based on block chain technology. In 2017
IEEE 19th Conference on Business Informatics (CBI) (Vol. 1, pp. 140-149). IEEE. 10.1109/CBI.2017.56
Nandy, T., Idris, M. Y. I. B., Noor, R. M., Kiah, L. M., Lun, L. S., Juma’at, N. B. A., ... Bhattacharyya, S. (2019). Review
on security of Internet of Things authentication mechanism. IEEE Access: Practical Innovations, Open Solutions, 7,
151054–151089.
Narendrakumar, S., Razaque, A., Patel, V., Almi’ani, M., Rizvi, S. S., & Hans, A. (2018). Token security for internet of
things. International Journal of Embedded Systems, 10(4), 334–343.
283
Compilation of References
Narudin, F. A., Feizollah, A., Anuar, N. B., & Gani, A. (2016). Evaluation of machine learning classifiers for mobile
malware detection. Soft Computing, 20(1), 343–357. doi:10.100700500-014-1511-6
Nataraj, L., Karthikeyan, S., Jacob, G., & Manjunath, B. S. (2011, July). Malware images: visualization and automatic
classification. In Proceedings of the 8th international symposium on visualization for cyber security (pp. 1-7). Academic
Press.
Naz, M., Al-zahrani, F. A., Khalid, R., Javaid, N., Qamar, A. M., Afzal, M. K., & Shafiq, M. (2019). A Secure Data
Sharing Platform Using Blockchain and Interplanetary File System. Sustainability, 11(24), 7054. doi:10.3390u11247054
Necla, B., & Ismail, E. (2012). WSNSec: A scalable data link layer security protocol forWSNs. Ad Hoc Networks, 10(1),
37–45. doi:10.1016/j.adhoc.2011.04.013
Neely, R. M., Piech, D. K., Santacruz, S. R., Maharbiz, M. M., & Carmena, J. M. (2018). Recent advances in neural
dust: Towards a neural interface platform. Current Opinion in Neurobiology, 50, 64–71. doi:10.1016/j.conb.2017.12.010
PMID:29331738
Nguyen, D. C., Pathirana, P. N., Ding, M., & Seneviratne, A. (2019). Blockchain for secure EHRs sharing of mobile
cloud based e-Health systems. IEEE Access: Practical Innovations, Open Solutions, 7, 66792–66806. doi:10.1109/AC-
CESS.2019.2917555
Nicholson, A., Webber, S., Dyer, S., Patel, T., & Janicke, H. (2012). SCADA security in the light of Cyber-Warfare.
Computers & Security, 31(4), 418–436. doi:10.1016/j.cose.2012.02.009
Noor, Z., Jung, L., Alsaadi, F., & Alghamdi, T. (2012). Wireless sensor network (WSN) routing security, reliability and
energy efficiency. J Appl Sci, 12(6), 593–59. doi:10.3923/jas.2012.593.597
Nurseitov, D., Serekov, A., Shintemirov, A., & Abibullaev, B. (2017). Design and evaluation of a P300-ERP based BCI
system for real-time control of a mobile robot. 2017 5th International Winter Conference on Brain-Computer Interface
(BCI), 1. 10.1109/IWW-BCI.2017.7858177
Nurse, J. R., Creese, S., & De Roure, D. (2017). Security risk assessment in Internet of Things systems. IT Professional,
19(5), 20–26.
O’Brien, S. A. (2016). Widespread cyberattack takes down sites world wide. Retrieved from https://money.cnn.
com/2016/10/21/technology/ddos-attack-popular-sites/index.html
Obaidat, M. S., Rana, S. P., Maitra, T., Giri, D., & Dutta, S. (2019). Biometric security and internet of things (IoT). In
Biometric-Based Physical and Cybersecurity Systems (pp. 477–509). Springer.
Ogonji, M. M., Okeyo, G., & Wafula, J. M. (2020). A survey on privacy and security of Internet of Things. Computer
Science Review, 38, 100312.
Ongtang, M., McLaughlin, S., Enck, W., & McDaniel, P. (2012). Semantically rich application‐centric security in An-
droid. Security and Communication Networks, 5(6), 658–673.
OpenViBE | Software for Brain Computer Interfaces and Real Time Neurosciences. (2020, December 10). OpenViBE.
http://openvibe.inria.fr/
Oracevic, A., Dilek, S., & Ozdemir, S. (2017, May). Security in internet of things: A survey. In 2017 International
Symposium on Networks, Computers and Communications (ISNCC) (pp. 1-6). IEEE. 10.1109/ISNCC.2017.8072001
284
Compilation of References
Ordóñez de Pablos, P., Almunawar, M. N., Chui, K. T., & Kaliannan, M. (Eds.). (2021). Handbook of Research on Ana-
lyzing IT Opportunities for Inclusive Digital Learning. IGI Global. doi:10.4018/978-1-7998-7184-2
Pais-Vieira, M., Lebedev, M., Kunicki, C., Wang, J., & Nicolelis, M. A. L. (2013). A Brain-to-Brain Interface for Real-
Time Sharing of Sensorimotor Information. Scientific Reports, 3(1), 1319. doi:10.1038rep01319 PMID:23448946
Pajouh, H., Dehghantanha, A., Parizi, R. M., Aledhari, M., & Karimipour, H. (2019). A survey on internet of things
security: Requirements, challenges, and solutions. Internet of Things, 100129.
Pang, Y. (2019). A signature-based assistant random oversampling method for malware detection. In 2019 18th IEEE
International conference on trust, security and privacy in computing and communications/13th IEEE international con-
ference on big data science and engineering (TrustCom/BigDataSE). IEEE. 10.1109/TrustCom/BigDataSE.2019.00042
Pan, S., Morris, T., & Adhikari, U. (2015). Developing a hybrid intrusion detection system using data mining for power
systems. IEEE Transactions on Smart Grid, 6(6), 3104–3113. doi:10.1109/TSG.2015.2409775
Pascanu, R., Stokes, J. W., Sanossian, H., Marinescu, M., & Thomas, A. (2015, April). Malware classification with
recurrent networks. In 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (pp.
1916-1920). IEEE. 10.1109/ICASSP.2015.7178304
Patwary, A. A. N., Fu, A., Battula, S. K., Naha, R. K., Garg, S., & Mahanti, A. (2020). FogAuthChain: A secure location-
based authentication scheme in fog computing environments using Blockchain. Computer Communications, 162, 212–224.
Peraković, D., Periša, M., Cvitić, I., & Zorić, P. (n.d.). Information and communication technologies for the society 5.0
environment. Academic Press.
Perakovic, D., Perisa, M., Cvitic, I., & Zoric, P. (2020). Identification of the relevant parameters for modeling the eco-
system elements in Industry 4.0. In 4th EAI International Conference on Management of Manufacturing Systems (pp.
111-123). Springer. 10.1007/978-3-030-34272-2_11
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdrán, A., García Clemente, F. J., & Cleary, F. (2021). Craft-
ing Adversarial Samples for Anomaly Detectors in Industrial Control Systems. In The 4th International Conference on
Emerging Data and Industry 4.0 (EDI40).
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdrán, A., & García Clemente, F. J. (2020). MADICS: A
Methodology for Anomaly Detection in Industrial Control Systems. Symmetry, 12(10), 1583. doi:10.3390ym12101583
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdran, A., García Clemente, F. J., Cadenas Sarmiento, C., Del
Canto Masa, C. J., & Méndez Nistal, R. (2019). On the generation of anomaly detection datasets in industrial control
systems. IEEE Access: Practical Innovations, Open Solutions, 7, 177460–177473. doi:10.1109/ACCESS.2019.2958284
Perales Gómez, Á. L., Fernández Maimó, L., Huertas Celdrán, A., García Clemente, F. J., Gil Pérez, M., & Martínez
Pérez, G. (2020). SafeMan: A unified framework to manage cybersecurity and safety in manufacturing industry. Soft-
ware, Practice & Experience.
Perrig, A., Szewczyk, R., Wen, V., Cullar, D., & Tygar, J. D. (2002). SPINS: Security protocols for sensor networks. Int
J Commun Comput Inform, 8(5), 521–534.
Pham, H. A., Le, T. K., & Le, T. V. (2019, September). Enhanced Security of IoT Data Sharing Management by Smart
Contracts and Blockchain. In 2019 19th International Symposium on Communications and Information Technologies
(ISCIT) (pp. 398-403). IEEE. 10.1109/ISCIT.2019.8905219
Pillitteri, V. Y., & Brewer, T. L. (2014). Guidelines for smart grid cybersecurity. Academic Press.
285
Compilation of References
Pinelli, M., Venturini, M., & Burgio, M. (2003, January). Statistical methodologies for reliability assessment of gas
turbine measurements. In Turbo Expo: Power for Land, Sea, and Air (Vol. 36851, pp. 787-793). 10.1115/GT2003-38407
Polich, J., & Heine, M. (1996). P300 topography and modality effects from a single-stimulus paradigm. Psychophysiol-
ogy, 33(6), 747–752. doi:10.1111/j.1469-8986.1996.tb02371.x PMID:8961797
Polich, J., Howard, L., & Starr, A. (1985). Effects of Age on the P300 Component of the Event-related Potential From
Auditory Stimuli: Peak Definition, Variation, and Measurement. Journal of Gerontology, 40(6), 721–726. doi:10.1093/
geronj/40.6.721 PMID:4056328
Prasse, P., Machlica, L., Pevný, T., Havelka, J., & Scheffer, T. (2017, September). Malware detection by analysing
encrypted network traffic with neural networks. In Joint European Conference on Machine Learning and Knowledge
Discovery in Databases (pp. 73-88). Springer. 10.1007/978-3-319-71246-8_5
Pycroft, L., & Aziz, T. Z. (2018). Security of implantable medical devices with wireless connections: The dangers of
cyber-attacks. Expert Review of Medical Devices, 15(6), 403–406. doi:10.1080/17434440.2018.1483235 PMID:29860880
Pycroft, L., Boccard, S. G., Owen, S. L. F., Stein, J. F., Fitzgerald, J. J., Green, A. L., & Aziz, T. Z. (2016). Brainjacking:
Implant Security Issues in Invasive Neuromodulation. World Neurosurgery, 92, 454–462. doi:10.1016/j.wneu.2016.05.010
PMID:27184896
Qin, T., & Chen, H. (2012). An Enhanced Scheme against Node Capture Attack using Hash-Chain for Wireless Sensor
Networks. Journal of Information Technology, 11(1), 102–109. doi:10.3923/itj.2012.102.109
Qu, X., Yang, L., Guo, K., Ma, L., Sun, M., Ke, M., & Li, M. (2021). A survey on the development of self-organizing maps
for unsupervised intrusion detection. Mobile Networks and Applications, 26(2), 808–829. doi:10.100711036-019-01353-0
Raff, E., Barker, J., Sylvester, J., Brandon, R., Catanzaro, B., & Nicholas, C. (2017). Malware detection by eating a
whole exe. arXiv preprint arXiv:1710.09435.
Rahman, M. S., Al Omar, A., Bhuiyan, M. Z. A., Basu, A., Kiyomoto, S., & Wang, G. (2020). Accountable cross-border
data sharing using blockchain under relaxed trust assumption. IEEE Transactions on Engineering Management.
Ramadan, R. A., & Vasilakos, A. V. (2017). Brain computer interface: Control signals review. Neurocomputing, 223,
26–44. doi:10.1016/j.neucom.2016.10.024
Ranjeetha, S., Renuga, N., & Sharmila, R. (2017) Secure zone routing protocol for MANET. International conference
on emerging trends in engineering, science and sustainable technology (ICETSST-2017), 67–76.
Rao, R. P. N. (2019). Towards neural co-processors for the brain: Combining decoding and encoding in brain–computer
interfaces. Current Opinion in Neurobiology, 55, 142–151. doi:10.1016/j.conb.2019.03.008 PMID:30954862
Rathi, N., Singla, R. & Tiwari, S. (2021). A novel approach for designing authentication system using a picture based
P300 speller. Cogn Neurodyn. doi:10.1007/s11571-021-09664-3
Rawat, D. B., Njilla, L., Kwiat, K., & Kamhoua, C. (2018, March). iShare: Blockchain-based privacy-aware multi-agent
information sharing games for cybersecurity. In 2018 International Conference on Computing, Networking and Com-
munications (ICNC) (pp. 425-431). IEEE. 10.1109/ICCNC.2018.8390264
Ray, B. R., Abawajy, J., & Chowdhury, M. (2014). Scalable RFID security framework and protocol supporting Internet
of Things. Computer Networks, 67, 89–103. doi:10.1016/j.comnet.2014.03.023
Ray, P. P. (2018). A survey on Internet of Things architectures. Journal of King Saud University-Computer and Informa-
tion Sciences, 30(3), 291–319.
286
Compilation of References
Razzaq, M. A., Gill, S. H., Qureshi, M. A., & Ullah, S. (2017). Security issues in the Internet of Things (IoT): A com-
prehensive study. International Journal of Advanced Computer Science and Applications, 8(6), 383.
Rege, A., & Bleiman, R. (2020, June). Ransomware Attacks Against Critical Infrastructure. In ECCWS 2020 20th Eu-
ropean Conference on Cyber Warfare and Security (p. 324). Academic Conferences and Publishing Limited. https://
www.risidata.com/
Ren, Z., Liu, X., Ye, R., & Zhang, T. (2017, July). Security and privacy on internet of things. In 2017 7th IEEE Interna-
tional Conference on Electronics Information and Emergency Communication (ICEIEC) (pp. 140-144). IEEE.
Rendell, D. (2019). Understanding the evolution of malware. Computer Fraud & Security, 2019(1), 17–19. doi:10.1016/
S1361-3723(19)30010-7
Ren, Z., Chen, G., & Lu, W. (2020). Malware visualization methods based on deep convolution neural networks. Mul-
timedia Tools and Applications, 79(15), 10975–10993. doi:10.100711042-019-08310-9
Rezende, E., Ruppert, G., Carvalho, T., Ramos, F., & De Geus, P. (2017, December). Malicious software classifica-
tion using transfer learning of resnet-50 deep neural network. In 2017 16th IEEE International Conference on Machine
Learning and Applications (ICMLA) (pp. 1011-1014). IEEE.
Richemond, P. H., & Guo, Y. (2019). Combining learning rate decay and weight decay with complexity gradient descent-
Part I. arXiv preprint arXiv:1902.02881.
Riel, A., Kreiner, C., Macher, G., & Messnarz, R. (2017). Integrated design for tackling safety and security challenges
of smart products and digital manufacturing. CIRP Annals, 66(1), 177–180. doi:10.1016/j.cirp.2017.04.037
Rizvi, S., Kurtz, A., Pfeffer, J., & Rizvi, M. (2018, August). Securing the internet of things (IoT): A security taxonomy for
IoT. In 2018 17th IEEE International Conference On Trust, Security And Privacy In Computing And Communications/12th
IEEE International Conference On Big Data Science And Engineering (TrustCom/BigDataSE) (pp. 163-168). IEEE.
Robertazzi, T. G. (2017). Software-defined networking. In Introduction to Computer Networking (pp. 81–87). Springer.
Roman, R., Najera, P., & Lopez, J. (2011). Securing the internet of things. Computer, 44(9), 51–58.
Ronen, R., Radu, M., Feuerstein, C., Yom-Tov, E., & Ahmadi, M. (2018). Microsoft malware classification challenge.
arXiv preprint arXiv:1802.10135.
Rose, G., Raghuram, P., Watson, S., & Wigley, E. (2021). Platform urbanism, smartphone applications and valuing data
in a smart city. Transactions of the Institute of British Geographers, 46(1), 59–72. doi:10.1111/tran.12400
Saboor, A., Gembler, F., Benda, M., Stawicki, P., Rezeika, A., Grichnik, R., & Volosyak, I. (2018). A Browser-Driven
SSVEP-Based BCI Web Speller. 2018 IEEE International Conference on Systems, Man, and Cybernetics (SMC), 625–630.
10.1109/SMC.2018.00115
Sadeeq, M. A., Zeebaree, S. R., Qashi, R., Ahmed, S. H., & Jacksi, K. (2018, October). Internet of Things security: a
survey. In 2018 International Conference on Advanced Science and Engineering (ICOASE) (pp. 162-166). IEEE.
Sadeghi, A. R., Wachsmann, C., & Waidner, M. (2015, June). Security and privacy challenges in industrial internet of
things. In 2015 52nd ACM/EDAC/IEEE Design Automation Conference (DAC) (pp. 1-6). IEEE.
Saha, H. N., Mandal, A., & Sinha, A. (2017, January). Recent trends in the Internet of Things. In 2017 IEEE 7th annual
computing and communication workshop and conference (CCWC) (pp. 1-4). IEEE.
Sahmim, S., & Gharsellaoui, H. (2017). Privacy and security in internet-based computing: cloud computing, internet
of things, cloud of things: a review. Procedia Computer Science, 112, 1516–1522. doi:10.1016/j.procs.2017.08.050
287
Compilation of References
Sain, M., Kang, Y. J., & Lee, H. J. (2017, February). Survey on security in Internet of Things: State of the art and chal-
lenges. In 2017 19th International conference on advanced communication technology (ICACT) (pp. 699-704). IEEE.
Salam, A. (2020). Internet of things in agricultural innovation and security. In Internet of Things for Sustainable Com-
munity Development (pp. 71–112). Springer.
Salhi, D. E., Tari, A., & Kechadi, M. T. (2021). Using Clustering for Forensics Analysis on Internet of Things. Inter-
national Journal of Software Science and Computational Intelligence, 13(1), 56–71. doi:10.4018/IJSSCI.2021010104
Salman, T., & Jain, R. (2019). A survey of protocols and standards for internet of things. arXiv preprint arXiv:1903.11549.
Samaila, M. G., Neto, M., Fernandes, D. A., Freire, M. M., & Inácio, P. R. (2017). Security challenges of the Internet
of Things. In Beyond the Internet of Things (pp. 53–82). Springer.
Samuel, O., Javaid, N., Awais, M., Ahmed, Z., Imran, M., & Guizani, M. (2019, July). A blockchain model for fair
data sharing in deregulated smart grids. In IEEE Global Communications Conference (GLOBCOM 2019). 10.1109/
GLOBECOM38437.2019.9013372
Sarma, A., Matos, A., Girao, J., & Aguiar, R. L. (2008). Virtual identity framework for telecom infrastructures. Wireless
Personal Communications, 45(4), 521–543.
Sathyanarayan, V. S., Kohli, P., & Bruhadeshwar, B. (2008). Signature generation and detection of malware families. In
Australasian Conference on Information Security and Privacy. Springer. 10.1007/978-3-540-70500-0_25
Saxe, J., & Berlin, K. (2015, October). Deep neural network based malware detection using two dimensional binary
program features. In 2015 10th International Conference on Malicious and Unwanted Software (MALWARE) (pp. 11-
20). IEEE. 10.1109/MALWARE.2015.7413680
Saxe, J., & Sanders, H. (2018). Malware Data Science: Attack Detection and Attribution. No Starch Press.
Sempreboni, D., & Viganò, L. (2018). Privacy, Security and Trust in the Internet of Neurons. arXiv.
Seo., D., Carmena, J. M., Rabaey, J. M., Alon, E., & Maharbiz, M. M. (2013). Neural Dust: An Ultrasonic, Low Power
Solution for Chronic Brain-Machine Interfaces. arXiv.
Sfar, A. R., Natalizio, E., Challal, Y., & Chtourou, Z. (2018). A roadmap for security challenges in the Internet of Things.
Digital Communications and Networks, 4(2), 118–137. doi:10.1016/j.dcan.2017.04.003
Shaila, K., Manjula, S. H., Thriveni, J., Venugopal, K. R., & Patnaik, L. M. (2011). Resilience against node capture
attack using asymmetric matrices in key predistribution scheme in wireless sensor networks. International Journal on
Computer Science and Engineering, 3(10), 3490.
Sha, K., Wei, W., Yang, T. A., Wang, Z., & Shi, W. (2018). On security challenges and open issues in Internet of Things.
Future Generation Computer Systems, 83, 326–337.
Shalyga, D., Filonov, P., & Lavrentyev, A. (2018). Anomaly detection for water treatment system based on neural network
with automatic architecture optimization. arXiv preprint arXiv:1807.07282.
Sharafaldin, I., Lashkari, A. H., & Ghorbani, A. A. (2018, January). Toward generating a new intrusion detection dataset
and intrusion traffic characterization. In ICISSp (pp. 108-116). doi:10.5220/0006639801080116
Sharmeen, S., Huda, S., Abawajy, J. H., Ismail, W. N., & Hassan, M. M. (2018). Malware threats and detection for
industrial mobile-IoT networks. IEEE Access: Practical Innovations, Open Solutions, 6, 15941–15957. doi:10.1109/
ACCESS.2018.2815660
288
Compilation of References
Shhadat, I., Hayajneh, A., & Al-Sharif, Z. A. (2020). The use of machine learning techniques to advance the detection
and classification of unknown malware. Procedia Computer Science, 170, 917–922. doi:10.1016/j.procs.2020.03.110
Shi, G., Zhang, J., Zhang, C., & Hu, J. (2020). A distributed parallel training method of deep belief networks. Soft Com-
puting, 24(17), 1–12. doi:10.100700500-020-04754-6
Shrestha, A. K., Deters, R., & Vassileva, J. (2019). User-controlled privacy-preserving user profile data sharing based
on blockchain. arXiv preprint arXiv:1909.05028.
Shrestha, A. K., & Vassileva, J. (2018, June). Blockchain-based research data sharing framework for incentivizing the
data owners. In International Conference on Blockchain (pp. 259-266). Springer. 10.1007/978-3-319-94478-4_19
Shukla, P. K., Goyal, S., Wadhvani, R., Rizvi, M. A., Sharma, P., & Tantubay, N. (2015). Finding robust assailant using
optimization functions (FiRAO-PG) in wireless sensor network. Mathematical Problems in Engineering, 2015, 2015.
doi:10.1155/2015/594345
Sial, A., Singh, A., & Mahanti, A. (2019). Detecting anomalous energy consumption using contextual analysis of smart
meter data. Wireless Networks, 1–18.
Sial, A., Singh, A., Mahanti, A., & Gong, M. (2018, April). Heuristics-Based Detection of Abnormal Energy Consump-
tion. In International Conference on Smart Grid Inspired Future Technologies (pp. 21-31). Springer.
Sicari, S., Rizzardi, A., Grieco, L. A., & Coen-Porisini, A. (2015). Security, privacy and trust in Internet of Things: The
road ahead. Computer Networks, 76, 146–164.
Siddiqui, M., Wang, M. C., & Lee, J. (2008). A survey of data mining techniques for malware detection using file features.
Proceedings of the 46th annual southeast regional conference. 10.1145/1593105.1593239
Si, H., Sun, C., Li, Y., Qiao, H., & Shi, L. (2019). IoT information sharing security mechanism based on blockchain
technology. Future Generation Computer Systems, 101, 1028–1040. doi:10.1016/j.future.2019.07.036
Sihwail, Omar, & Ariffin. (2018). A survey on malware analysis techniques: Static, dynamic, hybrid and memory analysis.
International Journal on Advanced Science, Engineering and Information Technology, 8(4-2), 1662.
Simon, A. J., Bernstein, A., Hess, T., Ashrafiuon, H., Devilbiss, D., & Verma, A. (2011). P1-112: A brain computer inter-
face to detect Alzheimer’s disease. Alzheimer’s & Dementia, 7(4S_Part_4), S145–S146. doi:10.1016/j.jalz.2011.05.391
Sinaović, H., & Mrdovic, S. (2017). Analysis of Mirai malicious software. In Proc. of SoftCOM 2017 (pp. 1-5). HR.
Singh, A., & Kumar, R. (2021). A Two-Phase Load Balancing Algorithm for Cloud Environment. International Journal
of Software Science and Computational Intelligence, 13(1), 38–55.
Singh, N., & Vardhan, M. (2019). Distributed ledger technology based property transaction system with support for iot
devices. International Journal of Cloud Applications and Computing, 9(2), 60–78.
Sklavos, N., & Zaharakis, I. D. (2016, November). Cryptography and security in internet of things (iots): Models,
schemes, and implementations. In 2016 8th IFIP International Conference on New Technologies, Mobility and Security
(NTMS) (pp. 1-2). IEEE.
Skobelev, P. O., & Borovik, S. Y. (2017). On the way from Industry 4.0 to Industry 5.0: from digital manufacturing to
digital society. Industry 4.0, 2(6), 307-311.
289
Compilation of References
Solis, D., & Vicens, R. (2017, October). Convolutional neural networks for classification of malware assembly code. In
Recent Advances in Artificial Intelligence Research and Development: Proceedings of the 20th International Confer-
ence of the Catalan Association for Artificial Intelligence, Deltebre, Terres de L’Ebre, Spain, October 25-27, 2017 (Vol.
300, p. 221). IOS Press.
SonicWall. (2021). Annual number of malware attacks worldwide from 2015 to 2020 (in billions). In Statista. Retrieved
9 August 2021, from https://www.statista.com/statistics/873097/malware-attacks-per-year-worldwide/
Sowndhararajan, K., Minju, K., Ponnuvel, D., Se, P., & Songmun, K. (2018). Application of the P300 Event-Related Po-
tential in the Diagnosis of Epilepsy Disorder: A Review. Scientia Pharmaceutica, 86(2), 10. doi:10.3390cipharm86020010
PMID:29587468
Stankovic, J. A. (2014). Research directions for the internet of things. IEEE Internet of Things Journal, 1(1), 3–9.
StatCounter Global Stats. (2020). Mobile & Tablet Android Version Market Share Worldwide. https://gs.statcounter.com/
android-version-market-share/mobile-tablet/worldwide
Stein, G., Chen, B., Wu, A. S., & Hua, K. A. (2005, March). Decision tree classifier for network intrusion detection with
GA-based feature selection. In Proceedings of the 43rd annual Southeast regional conference-Volume 2 (pp. 136-141).
10.1145/1167253.1167288
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based big data secure management in the fog over a 6G wire-
less network. IEEE Internet of Things Journal, 8(7), 5164–5171. doi:10.1109/JIOT.2020.3033131
Stergiou, C. L., Psannis, K. E., & Gupta, B. B. (2020). IoT-based Big Data secure management in the Fog over a 6G
Wireless Network. IEEE Internet of Things Journal.
Sumathi, M., & Sangeetha, S. (2020). Blockchain based sensitive attribute storage and access monitoring in banking
system. International Journal of Cloud Applications and Computing, 10(2), 77–92. doi:10.4018/IJCAC.2020040105
Sundaram, B. V., Ramnath, M., Prasanth, M., & Sundaram, V. (2015, March). Encryption and hash based security in
Internet of Things. In 2015 3rd International Conference on Signal Processing, Communication and Networking (IC-
SCN) (pp. 1-6). IEEE.
Sundararajan, K. (2017). Privacy and security issues in Brain Computer Interface [Unpublished master’s thesis]. Auck-
land University of Technology, Auckland, New Zealand.
Sun, W., Cai, Z., Li, Y., Liu, F., Fang, S., & Wang, G. (2018). Security and privacy in the medical internet of things: A
review. Security and Communication Networks.
Suto, L. (2010). Analyzing the accuracy and time costs of web application security scanners. Academic Press.
Tague, P., Slater, D., Rogers, J., & Poovendran, R. (2008, April). Vulnerability of network traffic under node capture
attacks using circuit theoretic analysis. In IEEE INFOCOM 2008-The 27th Conference on Computer Communications
(pp. 161-165). IEEE 10.1109/INFOCOM.2008.41
Taheri, A., & Lashkari. (2019). Extensible Android Malware Detection and Family Classification Using Network-Flows
and API-Calls. The IEEE (53rd) International Carnahan Conference on Security Technology.
Tahsien, S. M., Karimipour, H., & Spachos, P. (2020). Machine learning based solutions for security of Internet of Things
(IoT): A survey. Journal of Network and Computer Applications, 161, 102630.
290
Compilation of References
Takabi, H., Bhalotiya, A., & Alohaly, M. (2016). Brain Computer Interface (BCI) Applications: Privacy Threats and
Countermeasures. 2016 IEEE 2nd International Conference on Collaboration and Internet Computing (CIC), 102-111.
10.1109/CIC.2016.026
Takabi, H. (2016) Firewall for brain: Towards a privacy preserving ecosystem for BCI applications. IEEE Conference
on Communications and Network Security (CNS), 370-371. 10.1109/CNS.2016.7860516
Takase, T., Oyama, S., & Kurihara, M. (2018). Effective neural network training with adaptive learning rate based on
training loss. Neural Networks, 101, 68–78. doi:10.1016/j.neunet.2018.01.016 PMID:29494873
Tankard, C. (2015). The security issues of the Internet of Things. Computer Fraud & Security, 2015(9), 11–14.
Taormina, R., Galelli, S., Tippenhauer, N. O., Salomons, E., Ostfeld, A., Eliades, D. G., Aghashahi, M., Sundararajan,
R., Pourahmadi, M., Banks, M. K., Brentan, B. M., Campbell, E., Lima, G., Manzi, D., Ayala-Cabrera, D., Herrera,
M., Montalvo, I., Izquierdo, J., Luvizotto, E. Jr, ... Ohar, Z. (2018). Battle of the attack detection algorithms: Disclosing
cyber attacks on water distribution networks. Journal of Water Resources Planning and Management, 144(8), 04018048.
doi:10.1061/(ASCE)WR.1943-5452.0000969
Tao, Y., Xiangyang, X., Tonghui, L., & Leina, P. (2018). A secure routing of wireless sensor networks based on trust
evaluation model. Procedia Computer Science, 131, 1156–1163. doi:10.1016/j.procs.2018.04.289
Tavallaee, M., Bagheri, E., Lu, W., & Ghorbani, A. A. (2009, July). A detailed analysis of the KDD CUP 99 data set. In
2009 IEEE symposium on computational intelligence for security and defense applications (pp. 1-6). IEEE.
Technology markets: Software, Revenue by segment. (2021). In Statista. Retrieved 9 August 2021, from https://www.
statista.com/statistics/873097/malware-attacks-per-year-worldwide/
Tewari, A., & Gupta, B. B. (2017). A lightweight mutual authentication protocol based on elliptic curve cryptography
for IoT devices. International Journal of Advanced Intelligence Paradigms, 9(2-3), 111–121.
Tewari, A., & Gupta, B. B. (2020). Secure Timestamp-Based Mutual Authentication Protocol for IoT Devices Using RFID
Tags. International Journal on Semantic Web and Information Systems, 16(3), 20–34. doi:10.4018/IJSWIS.2020070102
Tewari, A., & Gupta, B. B. (2020). Security, privacy and trust of different layers in Internet-of-Things (IoTs) framework.
Future Generation Computer Systems, 108, 909–920.
Theodouli, A., Arakliotis, S., Moschou, K., Votis, K., & Tzovaras, D. (2018, August). On the design of a Blockchain-
based system to facilitate Healthcare Data Sharing. In 2018 17th IEEE International Conference on Trust, Security And
Privacy In Computing And Communications/12th IEEE International Conference On Big Data Science And Engineering
(TrustCom/BigDataSE) (pp. 1374-1379). IEEE. 10.1109/TrustCom/BigDataSE.2018.00190
Thwin, T. T., & Vasupongayya, S. (2018, August). Blockchain based secret-data sharing model for personal health record
system. In 2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)
(pp. 196-201). IEEE. 10.1109/ICAICTA.2018.8541296
Tyler, W. J., Sanguinetti, J. L., Fini, M., & Hool, N. (2017). Non-invasive neural stimulation. Micro- and Nanotechnology
Sensors. Systems, and Applications, IX, 280–290. doi:10.1117/12.2263175
Tzafestas, S. G. (2018). Ethics and law in the internet of things world. Smart Cities, 1(1), 98-120.
291
Compilation of References
Ucci, D., Aniello, L., & Baldoni, R. (2019). Survey of machine learning techniques for malware analysis. Computers &
Security, 81, 123–147. doi:10.1016/j.cose.2018.11.001
US-CERT. (2016). Heightened DDoS threat posed by Mirai and other botnets. Retrieved from https://www.us-cert.gov/
ncas/alerts/TA16-288A
Vaid, S., Singh, P., & Kaur, C. (2015). EEG Signal Analysis for BCI Interface: A Review. 2015 Fifth International
Conference on Advanced Computing & Communication Technologies, 143-147. 10.1109/ACCT.2015.72
Van Oorschot, P. C., & Smith, S. W. (2019). The internet of things: Security challenges. IEEE Security and Privacy,
17(5), 7–9.
Vardhana, M., Arunkumar, N., Lasrado, S., Abdulhay, E., & Ramirez-Gonzalez, G. (2018). Convolutional neural network
for bio-medical image segmentation with hardware acceleration. Cognitive Systems Research, 50, 10–14. doi:10.1016/j.
cogsys.2018.03.005
Vashi, S., Ram, J., Modi, J., Verma, S., & Prakash, C. (2017, February). Internet of Things (IoT): A vision, architectural
elements, and security issues. In 2017 international conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud)
(I-SMAC) (pp. 492-496). IEEE.
Vasighi, M., & Amini, H. (2017). A directed batch growing approach to enhance the topology preservation of self-
organizing map. Applied Soft Computing, 55, 424–435. doi:10.1016/j.asoc.2017.02.015
Vasilomanolakis, E., Daubert, J., Luthra, M., Gazis, V., Wiesmaier, A., & Kikiras, P. (2015, September). On the secu-
rity and privacy of Internet of Things architectures and systems. In 2015 International Workshop on Secure Internet of
Things (SIoT) (pp. 49-57). IEEE.
Vignali, G., Bottani, E., Guareschi, N., Di Donato, L., Ferraro, A., & Pirozzi, M. (2019, June). Development of a 4.0
industry application for increasing occupational safety: guidelines for a correct approach. In 2019 IEEE International
Conference on Engineering, Technology and Innovation (ICE/ITMC) (pp. 1-6). IEEE. 10.1109/ICE.2019.8792814
Vijayanand, R., Devaraj, D., & Kannapiran, B. (2018). Intrusion detection system for wireless mesh network using
multiple support vector machine classifiers with genetic-algorithm-based feature selection. Computers & Security, 77,
304–314. doi:10.1016/j.cose.2018.04.010
Vinayakumar, R., Alazab, M., Soman, K. P., Poornachandran, P., Al-Nemrat, A., & Venkatraman, S. (2019). Deep
learning approach for intelligent intrusion detection system. IEEE Access: Practical Innovations, Open Solutions, 7,
41525–41550. doi:10.1109/ACCESS.2019.2895334
Virat, M. S., Bindu, S. M., Aishwarya, B., Dhanush, B. N., & Kounte, M. R. (2018, May). Security and privacy chal-
lenges in internet of things. In 2018 2nd International Conference on Trends in Electronics and Informatics (ICOEI)
(pp. 454-460). IEEE.
Waldert, S. (2016). Invasive vs. Non-Invasive Neuronal Signals for Brain-Machine Interfaces: Will One Prevail? Frontiers
in Neuroscience, 10, 295. doi:10.3389/fnins.2016.00295 PMID:27445666
Wang, L., Liu, W., & Han, X. (2017, December). Blockchain-based government information resource sharing. In 2017
IEEE 23rd International Conference on Parallel and Distributed Systems (ICPADS) (pp. 804-809). IEEE. 10.1109/
ICPADS.2017.00112
292
Compilation of References
Wang, Z., & Wu, Q. (2019, October). Incentive for Historical Block Data Sharing in Blockchain. In 2019 IEEE 10th
Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON) (pp. 0913-0919). IEEE.
10.1109/IEMCON.2019.8936209
Wang, Z., Tian, Y., & Zhu, J. (2018, August). Data sharing and tracing scheme based on blockchain. In 2018 8th Inter-
national Conference on Logistics, Informatics and Service Sciences (LISS) (pp. 1-6). IEEE. 10.1109/LISS.2018.8593225
Wang, D., Vinson, R., Holmes, M., Seibel, G., Bechar, A., Nof, S., & Tao, Y. (2019). Early detection of tomato spotted
wilt virus by hyperspectral imaging and outlier removal auxiliary classifier generative adversarial nets (OR-AC-GAN).
Scientific Reports, 9(1), 1–14. doi:10.103841598-019-40066-y PMID:30867450
Wang, H., Li, Z., Li, Y., Gupta, B. B., & Choi, C. (2020). Visual saliency guided complex image retrieval. Pattern
Recognition Letters, 130, 64–72.
Wang, H., Peng, M. J., Miao, Z., Liu, Y. K., Ayodeji, A., & Hao, C. (2021). Remaining useful life prediction techniques
for electric valves based on convolution auto encoder and long short term memory. ISA Transactions, 108, 333–342.
doi:10.1016/j.isatra.2020.08.031 PMID:32891421
Wang, H., & Song, Y. (2018). Secure cloud-based EHR system using attribute-based cryptosystem and blockchain.
Journal of Medical Systems, 42(8), 152. doi:10.100710916-018-0994-6 PMID:29974270
Wang, H., Zhang, Y., Chen, K., Sui, G., Zhao, Y., & Huang, X. (2019). Functional broadcast encryption with applications
to data sharing for cloud storage. Information Sciences, 502, 109–124. doi:10.1016/j.ins.2019.06.028
Wang, N., Jiang, T., Li, W., & Lv, S. (2017). Physical-layer security in Internet of Things based on compressed sensing
and frequency selection. IET Communications, 11(9), 1431–1437.
Wang, W., Feng, Y., & Dai, W. (2018). Topic analysis of online reviews for two competitive products using latent
Dirichlet allocation. Electronic Commerce Research and Applications, 29, 142–156. doi:10.1016/j.elerap.2018.04.003
Wang, W., Li, Y., Wang, X., Liu, J., & Zhang, X. (2018). Detecting Android malicious apps and categorizing benign
apps with ensemble of classifiers. Future Generation Computer Systems, 78, 987–994. doi:10.1016/j.future.2017.01.019
Wang, W., Zhu, M., Zeng, X., Ye, X., & Sheng, Y. (2017, January). Malware traffic classification using convolutional
neural network for representation learning. In 2017 International Conference on Information Networking (ICOIN) (pp.
712-717). IEEE. 10.1109/ICOIN.2017.7899588
Wang, Y. R., Sun, G. D., & Jin, Q. (2020). Imbalanced sample fault diagnosis of rotating machinery using conditional varia-
tional auto-encoder generative adversarial network. Applied Soft Computing, 92, 106333. doi:10.1016/j.asoc.2020.106333
Wang, Y., Zhang, A., Zhang, P., & Wang, H. (2019). Cloud-Assisted EHR Sharing With Security and Privacy Preserva-
tion via Consortium Blockchain. IEEE Access: Practical Innovations, Open Solutions, 7, 136704–136719. doi:10.1109/
ACCESS.2019.2943153
Wang, Z., Zhou, C., & Liu, Y. (2017). Efficient hybrid detection of node replication attacks in mobile sensor networks.
Mobile Information Systems.
Weber, R. H. (2010). Internet of Things–New security and privacy challenges. Computer Law & Security Review, 26(1),
23–30.
293
Compilation of References
Wei, W., Yang, A. T., Shi, W., & Sha, K. (2016, October). Security in internet of things: Opportunities and challenges.
In 2016 International Conference on Identification, Information and Knowledge in the Internet of Things (IIKI) (pp.
512-518). IEEE.
Wirdatmadja, S. A., Barros, M. T., Koucheryavy, Y., Jornet, J. M., & Balasubramaniam, S. (2017). Wireless Optogenetic
Nanonetworks for Brain Stimulation: Device Model and Charging Protocols. IEEE Transactions on Nanobioscience,
16(8), 859–872. doi:10.1109/TNB.2017.2781150 PMID:29364130
Wortmann, F., & Flüchter, K. (2015). Internet of things. Business & Information Systems Engineering, 57(3), 221–224.
Wroblewski, G. (2013). General method of program code obfuscation. In The 5th Conference on Information and
Knowledge Technology. IEEE.
Wronka, E., Kaiser, J., & Coenen, A. M. (2008). The auditory P3 from passive and active three-stimulus oddball para-
digm. Acta Neurobiologiae Experimentalis, 68(3), 362–372. PMID:18668159
Wu, H., Schwab, S., & Peckham, R. L. (2008). U.S. Patent No. 7,424,744. Washington, DC: U.S. Patent and Trademark
Office.
Wu, A., Zhang, Y., Zheng, X., Guo, R., Zhao, Q., & Zheng, D. (2019). Efficient and privacy-preserving traceable attribute-
based encryption in blockchain. Annales des Télécommunications, 74(7-8), 401–411. doi:10.100712243-018-00699-y
Wu, S., Roberts, K., Datta, S., Du, J., Ji, Z., Si, Y., Soni, S., Wang, Q., Wei, Q., Xiang, Y., Zhao, B., & Xu, H. (2020).
Deep learning in clinical natural language processing: A methodical review. Journal of the American Medical Informat-
ics Association: JAMIA, 27(3), 457–470. doi:10.1093/jamia/ocz200 PMID:31794016
Xiang, W., Zhang, H., Cui, R., Chu, X., Li, K., & Zhou, W. (2018). Pavo: A rnn-based learned inverted index, supervised
or unsupervised? IEEE Access: Practical Innovations, Open Solutions, 7, 293–303. doi:10.1109/ACCESS.2018.2885350
Xie, J., Tang, H., Huang, T., Yu, F. R., Xie, R., Liu, J., & Liu, Y. (2019). A survey of blockchain technology applied
to smart cities: Research issues and challenges. IEEE Communications Surveys and Tutorials, 21(3), 2794–2830.
doi:10.1109/COMST.2019.2899617
Xin, Y., Kong, L., Liu, Z., Chen, Y., Li, Y., Zhu, H., Gao, M., Hou, H., & Wang, C. (2018). Machine learning and deep
learning methods for cybersecurity. IEEE Access: Practical Innovations, Open Solutions, 6, 35365–35381. doi:10.1109/
ACCESS.2018.2836950
Xuan, S., Zheng, L., Chung, I., Wang, W., Man, D., Du, X., & Guizani, M. (2020). An incentive mechanism for data
sharing based on blockchain with smart contracts. Computers & Electrical Engineering, 83, 106587. doi:10.1016/j.
compeleceng.2020.106587
Xu, L., Shah, N., Chen, L., Diallo, N., Gao, Z., Lu, Y., & Shi, W. (2017, April). Enabling the sharing economy: Privacy
respecting contract based on public blockchain. In Proceedings of the ACM Workshop on Blockchain, Cryptocurrencies
and Contracts (pp. 15-21). 10.1145/3055518.3055527
Yamaguchi, S. (2021). A Basic Command and Control Strategy in Botnet Defense System. In Proc. of IEEE ICCE 2021.
Academic Press.
Yamaguchi, S. (2020a). Botnet Defense System: Concept, Design, and Basic Strategy. Information, 11, 516.
Yamaguchi, S. (2020b). White-Hat Worm to Fight Malware and Its Evaluation by Agent-Oriented Petri Nets. Sensors
(Basel), 20, 556.
294
Compilation of References
Yamaguchi, S., Bin Ahmadon, M. A., & Ge, Q. W. (2016). Introduction of Petri Nets: Its Applications and Security Chal-
lenges. In B. Gupta, D. P. Agrawal, & S. Yamaguchi (Eds.), Handbook of Research on Modern Cryptographic Solutions
for Computer and Cyber Security (pp. 145–179). IGI Publishing.
Yamaguchi, S., & Gupta, B. (2019). Malware Threat in Internet of Things and Its Mitigation Analysis. In R. C. Joshi,
B. Gupta, D. P. Agrawal, & S. Yamaguchi (Eds.), Security, Privacy, and Forensics Issues in Big Data (pp. 363–379).
IGI Publishing.
Yamaguchi, S., & Gupta, B. (2021). Malware threat in Internet of Things and its mitigation analysis. In Research An-
thology on Combating Denial-of-Service Attacks (pp. 371–387). IGI Global. doi:10.4018/978-1-7998-5348-0.ch020
Yamaguchi, S., Tanaka, H., & Bin Ahmadon, M. A. (2020). Modeling and Evaluation of Mitigation Methods against IoT
Malware Mirai with Agent-Oriented Petri Net PN2. International Journal of Internet of Things and Cyber-Assurance,
1(3/4), 195–213.
Yang, K., Liu, D., Qu, Q., Sang, Y., & Lv, J. (2021). An automatic evaluation metric for Ancient-Modern Chinese trans-
lation. Neural Computing & Applications, 33(8), 3855–3867. doi:10.100700521-020-05216-8
Yang, T., Hakimian, S., & Schwartz, T. H. (2014). Intraoperative ElectroCorticoGraphy (ECog): Indications, techniques,
and utility in epilepsy surgery. Epileptic Disorders, 16(3), 271–279. doi:10.1684/epd.2014.0675 PMID:25204010
Yang, Y., Wu, L., Yin, G., Li, L., & Zhao, H. (2017). A survey on security and privacy issues in Internet-of-Things. IEEE
Internet of Things Journal, 4(5), 1250–1258.
Yaqoob, I., & Ahmed, E., ur Rehman, M. H., Ahmed, A. I. A., Al-garadi, M. A., Imran, M., & Guizani, M. (2017).
The rise of ransomware and emerging security challenges in the Internet of Things. Computer Networks, 129, 444–458.
Yin, C., Zhang, S., Wang, J., & Xiong, N. N. (2020). Anomaly detection based on convolutional recurrent autoencoder for
IoT time series. IEEE Transactions on Systems, Man, and Cybernetics. Systems, 1–11. doi:10.1109/TSMC.2020.2968516
Yousefi, A., & Jameii, S. M. (2017, May). Improving the security of internet of things using encryption algorithms. In
2017 International Conference on IoT and Application (ICIOT) (pp. 1-5). IEEE. 10.1109/ICIOTA.2017.8073627
Yousefi-Azar, M., Varadharajan, V., Hamey, L., & Tupakula, U. (2017). Autoencoder-based feature learning for cy-
ber security applications. 2017 International Joint Conference on Neural Networks (IJCNN), 3854-3861. 10.1109/
IJCNN.2017.7966342
Yousuf, O., & Mir, R. N. (2019). A survey on the internet of things security: State-of-art, architecture, issues and coun-
termeasures. Information & Computer Security.
Yuan, X., Li, L., & Wang, Y. (2019). Nonlinear dynamic soft sensor modeling with supervised long short-term memory
network. IEEE Transactions on Industrial Informatics, 16(5), 3168–3176. doi:10.1109/TII.2019.2902129
Yue, B., Fu, J., & Liang, J. (2018). Residual recurrent neural networks for learning sequential representations. Informa-
tion (Basel), 9(3), 56. doi:10.3390/info9030056
Yu, Y., Li, Y., Tian, J., & Liu, J. (2018). Blockchain-based solutions to security and privacy issues in the internet of
things. IEEE Wireless Communications, 25(6), 12–18.
Zander, T. O., Kothe, C., Jatzev, S., & Gaertner, M. (2010). Enhancing Human-Computer Interaction with Input from
Active and Passive Brain-Computer Interfaces. In D. Tan & A. Nijholt (Eds.), Brain-Computer Interfaces. Human-
Computer Interaction Series. Springer. doi:10.1007/978-1-84996-272-8_11
295
Compilation of References
Zarras, A. (2014). Automated generation of models for fast and precise detection of HTTP-based malware. In 2014
Twelfth Annual International Conference on Privacy, Security and Trust. IEEE. 10.1109/PST.2014.6890946
Zavrak, S., & Iskefiyeli, M. (2020). Anomaly-based intrusion detection from network flow features using variational au-
toencoder. IEEE Access: Practical Innovations, Open Solutions, 8, 108346–108358. doi:10.1109/ACCESS.2020.3001350
Zekri, El Kafhali, Aboutabit, & Saadi. (2017). DDoS attack detection using machine learning techniques in cloud com-
puting environments. . doi:10.1109/CloudTech.2017.8284731
Zhang, Wu, & Cao. (2008). A Secret Sharing-Based Key Management in Hierarchical Wireless Sensor Network. In
Securing wireless sensor networks: A survey. IEEE Communications Surveys and Tutorials.
Zhang, X., Zhao, J., & LeCun, Y. (2015). Character-level convolutional networks for text classification. arXiv preprint
arXiv:1509.01626.
Zhang, Y., He, D., & Choo, K. K. R. (2018). BaDS: Blockchain-based architecture for data sharing with ABS and CP-
ABE in IoT. Wireless Communications and Mobile Computing, 2018, 2018. doi:10.1155/2018/2783658
Zhang, H., Sindagi, V., & Patel, V. M. (2019). Image de-raining using a conditional generative adversarial network. IEEE
Transactions on Circuits and Systems for Video Technology, 30(11), 3943–3956. doi:10.1109/TCSVT.2019.2920407
Zhang, Y., Wu, C., Cao, J., & Li, X. (2013). A secret sharing-based key management in a hierarchical wireless sensor
network. International Journal of Distributed Sensor Networks, 2013(6), 1–7. doi:10.1155/2013/406061
Zhang, Z., Sun, R., Zhao, C., Wang, J., Chang, C. K., & Gupta, B. B. (2017). CyVOD: A novel trinity multimedia social
network scheme. Multimedia Tools and Applications, 76(18), 18513–18529. doi:10.100711042-016-4162-z
Zheng, H., Zhang, Z., Jiang, S., Yan, B., Shi, X., Xie, Y., Huang, X., Yu, Z., Liu, H., Weng, S., Nurmikko, A., Zhang, Y.,
Peng, H., Xu, W., & Zhang, J. (2019). A shape-memory and spiral light-emitting device for precise multisite stimulation
of nerve bundles. Nature Communications, 10(1), 2790. doi:10.103841467-019-10418-3 PMID:31243276
Zhengwang, Y., Wen, T., Song, X., Liu, Z., & Fu, C. (2017). An efficient dynamic trust evaluation model for wireless
sensor networks. Journal of Sensors, 2017, 1–16.
Zhioua, S. (2013, July). The middle east under malware attack dissecting cyber weapons. In 2013 IEEE 33rd International
Conference on Distributed Computing Systems Workshops (pp. 11-16). IEEE. 10.1109/ICDCSW.2013.30
Zhu, Q., Bushnell, L., & Başar, T. (2012, December). Game-theoretic analysis of node capture and cloning attack with
multiple attackers in wireless sensor networks. In 2012 IEEE 51st IEEE Conference on Decision and Control (CDC)
(pp. 3404-3411). IEEE. 10.1109/CDC.2012.6426481
Zizzo, G., Hankin, C., Maffeis, S., & Jones, K. (2019). Intrusion detection for industrial control systems: Evaluation
analysis and adversarial attacks. arXiv preprint arXiv:1911.04278.
296
297
Brij B. Gupta received a PhD degree from Indian Institute of Technology Roorkee, India in the area
of Information and Cyber Security. In 2009, he was selected for the Canadian Commonwealth Scholar-
ship award. He has published more than 380 research papers in International Journals and Conferences
of high repute including IEEE, Elsevier, ACM, Springer, Wiley, Taylor & Francis, Inderscience, etc.
Dr. Gupta also received the Sir Visvesvaraya Young Faculty Research Fellowship Award in 2017, Best
Faculty Awards in 2018 and 2019 from the Ministry of Electronics and Information Technology, GoI
and NIT Kurukshetra, respectively. He also received the 2020 and 2021 ICT Express Elsevier Best
Reviewer Award. His Google scholar H-index is 57 (including i10 index: 187, Researchgate RG Score:
45.79) with over 11000 citations for his work. In addition, he is serving as a distinguished lecturer for
IEEE Consumer Technology Society and also included in the list of Top 2% Scientists in the world which
is created by the researchers from Stanford University. He is also working as principal investigator of
various R&D projects sponsored by various government funding agencies. He is serving/served as As-
sociate editor (or Guest Editor) of IEEE TII, IEEE Access, IEEE IoTM, ACM TOIT, IEEE IoT, FGCS,
COMCOM, etc. Moreover, Dr. Gupta is also leading International Journal of Cloud Applications and
Computing (IJCAC), IGI Global, USA and IJSSCI, and IJSWIS, IGI Global as Editor-in-Chief. He has
also served as Technical program committee (TPC) member or in other capacities of more than 100
International conferences worldwide. Dr. Gupta received outstanding paper awards in both regular and
student categories in 5th IEEE Global Conference on Consumer Electronics (GCCE) in Kyoto, Japan
during Oct. 7-10, 2016. Dr. Gupta is senior member of IEEE, ACM, Life Member, International As-
sociation of Engineers (IAENG), Life Member, IACSIT. He also worked as a post-doctoral research
fellow in UNB, Canada. He was also visiting researcher with Yamaguchi University, Japan, with Deakin
University, Australia and with Swinburne University of Technology, Australia during January, 2015 and
2018, July 2017, and Mar-Apr. 2018, respectively. Moreover, he was also visiting Professor in University
of Murcia, Spain in Jun.-Jul., 2018. Additionally, he was visiting professor with Temple university, USA
and Staffordshire University, UK during June, 2019 and July 2020, respectively. At present, Dr. Gupta
is group lead, Information and Cyber security, Programme coordinator, and Assistant Professor in the
Department of Computer Engineering, National Institute of Technology Kurukshetra India. He is also
working as honorary Associate Professor with Macquarie University, Australia and Adjunct Associate
Professor with Asia University, Taiwan, respectively. His research interests include Information security,
Cyber Security, Cloud Computing, Web security, Mobile/Smartphone, Intrusion detection, IoT Security,
AI, Social media, Computer networks and Phishing.
***
About the Contributors
Kwok Tai Chui received the B.Eng. degree in Electronic and Communication Engineering – Busi-
ness Intelligence Minor, with first-class honor, and Ph.D. degree in Electronic Engineering from City
University of Hong Kong. He was the recipient of international awards in several IEEE events. For
instance, he received the 2nd Prize Award (Postgraduate Category) of 2014 IEEE Region 10 Student
Paper Contest, and Best Paper Award in IEEE The International Conference on Consumer Electronics-
China, in both 2014 and 2015. He has industry experience as a Senior Data Scientist with an Internet
of Things (IoT) company. He is currently an Assistant Professor of the Department of Technology. He
is specialized in artificial intelligence and data science/analytics for smart city applications including
healthcare, smart grid, transportation, and education. He has served as Managing Editor of International
Journal on Semantic Web and Information Systems, Associate Editors, Guest Editors, and Editorial
Board Members for several ESCI/SCI journals.
Félix J. García Clemente received the Ph.D. degree in computer science from the University of
Murcia, in 2006. He is currently an Associate Professor with the Department of Computer Engineer-
ing, University of Murcia. His research interests include cybersecurity and management of distributed
communication networks. He is the coauthor of over 100 scientific publications and an active member
on different national and international research projects.
Francisco José García-Peñalvo received his bachelor’s degree in computing from the University
of Valladolid (Spain), and his PhD. degree from the University of Salamanca, where he is currently the
Head of the Research Group in Interaction and e-Learning (GRIAL). His main research interests focus on
eLearning, computers and education and digital ecosystems. He is the Editor in Chief of the Education
in the Knowledge Society journal and the Journal of Information Technology Research. He coordinates
the Doctoral Program in Education in the Knowledge Society.
Govind P. Gupta received his Ph.D. degree from Indian Institute of Technology, Roorkee, India,
in 2014. He is presently working as Assistant Professor in the Department of Information Technology
at National Institute of Technology, Raipur, India. His current research interests include Internet of
Things, Software-defined Networking, Network Security, Blockchain Application development for IoT,
Enterprise Blockchain. He has published more than 50 research articles in peer reviewed Journals and
reputed Conference proceedings. He is a professional member of the IEEE and ACM.
Alberto Huertas Celdrán received the M.Sc. and Ph.D. degrees in computer science from the Uni-
versity of Murcia, Spain. He is a Postdoctoral Researcher with the Department of Informatics, University
of Zurich. His scientific interests include medical cyber-physical systems, brain–computer interfaces,
cybersecurity, trust, data privacy, continuous authentication, semantic technology, context-aware systems,
and computer networks.
298
About the Contributors
Mavneet Kaur is a Senior student at NIT Kurukshetra, doing her research in information security
under the guidance of Professor BB Gupta.
Anish Khan received a B.Tech. degree in 2018, in Electronics and Communication Engineering,
and M.Tech. degree in 2020, in Electronics and Communication Engineering both from the Kurukshetra
University, Kurukshetra. His field of interest is Wireless Sensor Networks and Internet of Things (IoT).
He published 2 international conference papers and 1 book chapter in the Springer series. He holds IEEE
Student Membership and IEEE Consumer Technology Society Membership.
Sanjay Kumar has completed his B.E. with specialization in Computer Science and Engineering from
MNREC(now MNNIT) Allahabad, M. Tech. in Computer Science and Engineering from RGPV Bhopal
and Ph.D. in Information Technology from NIT Raipur. He is having 14 years of experience in academia
and 2 years of experience in industry. He is life member of IETE and CSI. He is senior member of IEEE.
Wen Liu received the B.Sc. degree (Hons.) in information and computing science from the Depart-
ment of Mathematics, Wuhan University of Technology, Wuhan, China, in 2009, and the Ph.D. degree
in mathematical imaging from The Chinese University of Hong Kong, Hong Kong, in 2015. He was a
Visiting Scholar with the College of Computer Science and Software Engineering, Shenzhen Univer-
sity, Shenzhen, China, and also with the Agency for Science, Technology, and Research (A*STAR),
Singapore. He is currently an Associate Professor with the School of Navigation, Wuhan University of
Technology. His research interests include computer vision, trajectory data mining, machine learning, and
computational navigation sciences. He is an Associate Editor of the International Journal on Semantic
Web and Information Systems.
Sergio López Bernal received the B.Sc. and M.Sc. degrees in computer science from the University
of Murcia, and the M.Sc. degree in architecture and engineering for the IoT from IMT Atlantique, France.
He is currently pursuing the Ph.D. degree with the University of Murcia. His research interests include
ICT security on brain–computer interfaces and network and information security.
Miltiadis D. Lytras is Associate Professor at Deree- ACG with a research focus on semantic web,
knowledge management and e-learning, with more than 100 publications. He has co-edited more than 45
special issues in International Journals (IEEE Transaction on Knowledge and Data Engineering, IEEE
Internet Computing, IEEE Transactions on Education, Computers in Human Behaviour, Interactive
Learning Environments, Journal of Knowledge Management, Journal of Computer Assisted Learning)
and has authored/[co-]edited 42 books [e.g. Open Source for Knowledge and Learning Management,
Ubiquitous and Pervasive Knowledge Management, Intelligent Learning Infrastructures for Knowledge
Intensive Organizations, Semantic Web Based Information Systems, China Information Technology
Handbook, Real World Applications of Semantic Web and Ontologies, Web 2.0: The Business Model,
etc]. He (has) serves(ed) as the (Co) Editor in Chief of 8 international journals (e.g. International Journal
on Semantic Web and Information Systems, International Journal of Knowledge Society Research, Inter-
national Journal of Knowledge and Learning, International Journal of Technology Enhanced Learning).
299
About the Contributors
Enrique Tomás Martínez Beltrán is working towards a Ph.D. in Computer Science at the University
of Murcia, Spain. He obtained a B.Sc. degree in Information and Communication Technologies and an
M.Sc. degree in New Technologies, specializing in information security, networks, and telematics. His
research interests include cybersecurity, Brain-Computer Interfaces, IoT, and Artificial Intelligence ap-
plied to different fields using Machine Learning and Deep Learning techniques.
Gregorio Martínez Pérez received the M.Sc. and Ph.D. degrees in computer science from the
University of Murcia, Spain. He is currently a Full Professor with the Department of Information and
Communications Engineering, University of Murcia. His research interests include cybersecurity, pri-
vacy, and networking, working on different national and European IST research projects on these topics.
Dragan Peraković received a B.Sc. degree in 1995, an M.Sc. degree in 2003, and a PhD in 2005, all
at the University of Zagreb, Croatia, EU. Dragan is Head of the Department for Information and Com-
munication Traffic and Head of Chair of Information Communication Systems and Services Manage-
ment, all at the Faculty of Transport and Traffic Sciences, University of Zagreb, where he is currently a
full professor. Dragan is visiting professor at the University of Mostar, Faculty of Science and Education
Sciences, Mostar / Bosnia and Herzegovina. Area of scientific interests and activities is modelling of
innovative communication ecosystems in the environment of the TELCO, transport system (ITS) and
Industry 4.0; AI & ML in cybersecurity, DDoS, Internet of Things; AI in e-forensic of communication
ecosystems (terminal devices/services); design and development of new innovative services and modules.
Ángel Luis Perales Gómez received the Ph.D. degree in Computer Science from the University of
Murcia (Spain). He is currently a postdoctoral researcher with the Department of Computer Engineering,
University of Murcia. His research interests include deep learning, cybersecurity of industrial control
systems, and the security of distributed communication networks.
Mario Quiles Pérez is a B.Eng. student in Computer Engineering at the University of Murcia (UMU).
Mention in Information and Communication Technologies. Currently in his last academic year, he is
investigating the possible attacks that could be carried out on brain-machine interfaces. Interested and
initiated in cybersecurity, he develops training scenarios for the detection of adversarial attacks on internal
networks. In the future, he hopes to dedicate himself fully to the world of cybersecurity.
300
About the Contributors
Chien-wen Shen is a Professor in the Department of Business Administration and the Division Di-
rector of the University Social Responsibility Office at the National Central University (NCU). He also
serves as the Director of the Yunus Social Business Center and the Center for Media and Social Impact
at NCU and the Executive Director of the Social Impact Institute of Taiwan. He received his M.S. and
Ph.D. in Industrial Engineering and Management Sciences from Northwestern University. His research
interests include big data analytics, information management, supply chain management, decision sci-
ence, and social business.
Santosh Kumar Smmarwar is working towards a Ph.D. degree in Information Technology from,
National Institute of Technology, Raipur, India. He earned his Ph.D. scholarship position as a talented
student in 2019. He received B.E. with a specialization in Computer Science and Engineering in 2008
from SATI Vidisha(RGPV Bhopal) and an M.S. with a specialization in Cyber Law and Information
Security from the Indian Institute of Information Technology, Allahabad, India, in 2011. His research
interests include Cybersecurity and Network security areas like intrusion detection systems and malware
detection systems, Internet of Things, Machine learning, Deep learning.
Shingo Yamaguchi is a Professor in the Graduate School of Sciences and Technology for Innovation
Yamaguchi University, Japan. He received the B.E., M.E., and D.E. degrees from Yamaguchi University,
Japan, in 1992, 1994, and 2002. He was a Visiting Scholar at the University of Illinois at Chicago, US,
in 2007. He is currently the Director of Information and Data Science Education Center, Yamaguchi
University. His research interests include AI, IoT, big data, and cybersecurity. He was the Executive
Conference Chair of IEEE ICCE 2021. He is a Member of the Board of Governors of IEEE Consumer
Technology Society. He is also the Editor-in-Chief of IEEE Consumer Electronics Magazine. He is a
Senior Member of IEEE and IEICE.
301
302
Index
A clustering 5-6, 16, 29, 35, 69, 78, 94, 96-98, 110, 172
collusion resistant 235
Actuator 93 confidentiality 21, 95-96, 112, 117, 158, 177-178,
adversary 219, 223-225, 235 186, 189-190, 200, 206-207, 209, 228, 236, 238,
anomaly detection 19-20, 27, 29-30, 39, 65, 68, 74, 244-247, 249-251, 254
78, 83-85, 89-93, 128 convolution technique 94, 96, 98-100, 102-103, 106
Anonymity 159, 199, 201-202 cyber threat 19, 22, 29, 81, 87, 205
attacks 1-3, 12, 15, 17, 20-24, 28, 42, 44, 48, 57, 60- cybersecurity 19, 21, 70, 74, 77-79, 83, 85, 87-89, 92-
61, 71, 74, 76-77, 79-85, 90, 92-93, 95-96, 115, 93, 127, 129, 145, 171, 176-179, 181, 193-194,
117, 128-131, 145-146, 148-149, 153, 158-159, 205, 216, 236, 238-240, 244, 246-253, 257
161-164, 166, 178-179, 181-182, 186, 188, 193-
194, 196, 199, 205, 219-223, 227, 229, 231-235, D
238-240, 244-251, 254, 257
authentication 21-22, 40, 58, 67, 72, 94, 97, 113-115, data 1-5, 7-8, 11-17, 20-26, 28, 30-31, 33-34, 37,
158, 162-163, 165, 169, 171, 174, 177, 179-180, 39, 41-47, 49-50, 52-56, 58-60, 62-64, 66, 68,
194-197, 199-200, 202, 207-208, 213, 239, 243- 71, 75-76, 78, 80-85, 87, 89-96, 98-100, 103,
244, 250 106, 109-110, 112-114, 121, 125, 128, 147-150,
availability 59, 78, 148, 158, 200, 206-207, 209, 236, 154-163, 168, 170, 172-173, 176-181, 184-186,
238, 244-246, 248-251, 254 188-191, 193-194, 198-221, 228-230, 236-240,
242, 244-251, 253-254, 256
B DDoS 3, 15, 18, 20-21, 38, 44, 56, 68-69, 93, 127-131,
145-147, 149, 153, 166-167, 171, 209, 234
backward secrecy 113-114, 116, 119, 122, 126 deep learning 5, 7, 15-17, 20-22, 27, 30, 39, 41, 47-48,
BCI 176-186, 188-196, 198, 236-240, 242-255, 258- 51, 55, 57, 59-63, 65-72, 74, 78, 84, 92-93, 163-164
259 distributed ledger 173, 199-201, 206, 209, 213
big data 16-17, 21-23, 39, 59, 64, 68, 71, 110, 125,
147, 150, 163, 172-173, 199, 211, 217 E
blockchain technology 125, 199-202, 204-206, 208-
210, 214-218 Electroencephalogram 176
botnet 29, 57, 79, 86, 89, 93, 127-135, 139, 144-147, 158 Electroencephalographic Signal 198
Brain-Computer Interface 178, 194-196, 198, 236, Event-Related Potential 176, 196-198, 258
254-257
F
C
feature engineering 7, 41
classification 5-6, 16-17, 22, 38, 47, 49, 55-58, 62, features 1, 4, 6-13, 15-17, 22, 44-46, 48-49, 53, 55, 58,
65-66, 68-72, 84, 89, 116, 145, 166, 191, 195- 63, 73, 83-85, 95, 153, 163, 204, 210, 213, 243
196, 201, 236, 238-240, 242-245, 247-248, 250, forward secrecy 113-116, 124, 126
254, 256 framework 21-23, 33, 37, 39, 43, 45, 49-50, 52, 54,
Index
63, 79, 87-90, 92, 110, 125, 150, 160-162, 164, 76, 79-82, 88, 93, 127-128, 130-132, 145-147,
166-167, 169-170, 173-174, 176, 178, 181-185, 161, 176, 178, 185-186, 188-190, 193-194, 236,
188-190, 192-194, 198, 204-210, 214, 217 238-239, 244, 251, 254, 256
malware analysis 1, 3-5, 7-8, 10, 12-15, 17, 20, 41,
G 43-45, 52
malware detection 3-6, 8, 12-13, 15-17, 21, 33-34,
group communications 112, 117, 125 38-39, 41-42, 44, 46-48, 52, 54-55, 57-61, 65-71,
74, 146, 189
H multi-agent system 127, 136, 144
Multimodal learning 41
hash function 119, 188, 198
Human Machine Interface 93 N
I network security 19-20, 27-28, 34, 41, 65, 257-258
node capture attack 112, 219-221, 223-225, 227,
Industrial applications 28, 37, 74 232-235
Industrial Control System 83, 89, 91, 93
Industrial Cybersecurity 74 P
integrity 21, 33, 95-96, 112, 114, 117, 158, 160, 176-
178, 181, 186, 188, 190, 198, 200, 202, 206-209, P300 176-184, 186, 189-194, 196-198, 239, 243,
214, 236, 238-239, 244-246, 248-251, 254 255-256, 258
Internet of Things (IoT) 15, 17, 20-21, 38-39, 57, 59, P300 ERP 198
66, 70-71, 75, 89, 110, 127-128, 145-151, 153, Petri net 127, 136, 144, 147
155, 157, 163-175, 200, 203, 210, 213, 216-217, privacy 16-18, 20-21, 42, 56, 90-91, 110, 114, 125,
220-221 147-149, 158, 163-166, 168-175, 177, 181, 189,
intrusion detection 19-23, 27-31, 33-40, 51, 66, 71, 195, 197, 199-200, 202, 205-210, 213-218, 220,
73-74, 77, 90-93, 128, 146 233, 239, 255, 257-258
intrusion detection system 20, 30-31, 38-40, 90-93, Programmable Logic Controller 93
128, 146
R
K
ransomware 3, 8, 20, 46, 51-52, 60, 76, 82, 92-93,
key independence 113, 126, 219 149, 175
key management 110-118, 123-126, 219, 233-235 rekeying 112-119, 121-123, 125-126
key management schemes 112, 115-116, 118, 123- Resistance Against Node Capture 219, 235
125, 219 risk 2, 19, 45, 52, 68, 86-88, 90, 127, 153, 162, 171, 204-
key predistribution 110, 126, 227-228, 233-235 206, 210, 236, 238, 244-246, 248-250, 253-254
L S
latent Dirichlet allocation 19, 23, 38, 40 safety 44, 53, 74, 79, 85-90, 92, 113-114, 158, 181,
LEACH 94, 97-98 195, 209, 236, 238, 244-251, 254
Logical Key Hierarchy 112, 115-119, 121, 125 SCADA 75, 81-82, 85-86, 89, 91, 93
secure multicast 112, 116, 119, 125-126
M security 2, 4, 8, 10, 14-22, 27-29, 31, 33-35, 37-42,
44, 46-47, 50, 55-57, 65, 67-72, 76, 79, 81-83,
machine learning 1-8, 10-11, 14-18, 20-23, 25, 28, 36, 87-92, 94-100, 103-106, 109-110, 112-115, 118,
38-39, 41-42, 46-48, 51-60, 69-71, 74, 78, 84, 123-125, 127, 145-149, 153, 157-160, 162-178,
89-90, 93, 110, 124, 166, 174, 210, 234 181, 186, 188-190, 192-197, 199-202, 204-211,
malicious node 94, 96-97, 99, 102, 106-108, 159 213-217, 220-221, 223-224, 233-234, 239, 244,
malware 1-23, 29-30, 33-34, 38-39, 41-61, 65-72, 74, 254-258
303
Index
sensor 21, 57, 70, 73, 78, 87, 93-99, 103, 106, 110-111, V
115, 123-126, 148-149, 154, 159, 169, 209-210,
214, 219-225, 233-235 visual stimuli 176, 178, 181, 183-184, 194, 198, 239,
signature 1, 3-4, 9, 14-15, 17, 46, 66, 69, 74, 95, 205- 244
207, 209, 215 vulnerability 52-53, 57, 81-82, 85-86, 128-129, 176,
smart contract 199, 202-204, 206-210 221, 223-224, 227, 235-236
Supervisory Control and Data Acquisition 75, 89, 93
systematic review 59, 61, 65 W
T Wireless Sensor Network (WSN) 96, 103, 106, 110-
111, 123, 125, 148, 220, 222, 233-234
topic modeling 19-21, 23, 37 worm 20, 81-82, 90-91, 127, 129-132, 134, 136-139,
141-147
304