The Convergence of

Internet of Things
and Cloud for Smart
Computing
Intelligent Signal Processing and Data
Analysis

SERIES EDITOR
Nilanjan Dey, Department of Information Technology, Techno India
College of Technology, Kolkata, India
Proposals for the series should be sent directly to one of the series
editors above, or submitted to:
Chapman & Hall/​CRC
Taylor and Francis Group
3 Park Square, Milton Park
Abingdon, OX14 4RN, UK

Translational Bioinformatics Applications in Healthcare

Khalid Raza, Nilanjan Dey

VANET
Challenges and Opportunities
Sonali P. Botkar, Sachin P. Godse, Parikshit N. Mahalle and
Gitanjali Shinde

Translational Bioinformatics Applications in Healthcare

Khalid Raza, Nilanjan Dey

COVID-​19 Public Health Measures

An Augmented Reality Perspective
Nuzhat F. Shaikh, Ajinkya Kunjir, Juveriya Shaikh,
Parikshit Narendra Mahalle

The Convergence of Internet of Things and Cloud for Smart

Computing
Parikshit N. Mahalle, Nancy Ambritta P., Gitanjali Rahul Shinde
and Arvind Vinayak Deshpande

For more information about this series, please visit: www.routledge.

com/​Intelligent-​Signal-​Processing-​and-​Data-​Analysis/​book-​series/​
INSPDA
 The Convergence of
Internet of Things
and Cloud for Smart
Computing

Parikshit N. Mahalle, Nancy Ambritta P.,

Gitanjali Rahul Shinde and Arvind Vinayak Deshpande
First edition published 2022
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-​2742
and by CRC Press
2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
© 2022 Parikshit N. Mahalle, Nancy Ambritta P., Gitanjali Rahul Shinde
and Arvind Vinayak Deshpande
CRC Press is an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the
author and publisher cannot assume responsibility for the validity of all materials or
the consequences of their use. The authors and publishers have attempted to trace
the copyright holders of all material reproduced in this publication and apologize to
copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted,
reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other
means, now known or hereafter invented, including photocopying, microfilming, and
recording, or in any information storage or retrieval system, without written permission
from the publishers.
For permission to photocopy or use material electronically from this work, access
www.copyright.com or contact the Copyright Clearance Center, Inc. (CCC),
222 Rosewood Drive, Danvers, MA 01923, 978-​750-​8400. For works that are not
available on CCC please contact [email protected]
Trademark notice: Product or corporate names may be trademarks or registered trademarks
and are used only for identification and explanation without intent to infringe.
Library of Congress Cataloging‑in‑Publication Data
A catalog record has been requested for this book
ISBN: 978-​1-​032-​03804-​9 (hbk)
ISBN: 978-​1-​032-​03805-​6 (pbk)
ISBN: 978-​1-​003-​18909-​1 (ebk)
Typeset in Times
by Newgen Publishing UK
 CO N T EN T S

Authors ix

Preface xii

Acknowledgments xv

1 Introduction 1
1.1 Overview of IoT 1
1.1.1 Technical Building Blocks 2
1.1.2 IoT in Business 5
1.2 IoT Architecture: Layered Perspective 15
1.3 Smart Computing 16
1.4 IoT Design: Issues and Challenges 17
1.4.1 Design Issues 17
1.4.2 Challenges 19

2 Internet of Things Application Development 23

2.1 Application Development Phases 23
2.1.1 Principles 23
2.1.2 C Model 24
2.2 Wireless Technologies for IoT 26
2.2.1 Bluetooth, Bluetooth Low Energy, and
Bluetooth 5 26
2.2.2 Zigbee 27
2.2.3 WiFi 27
2.2.4 6LowPAN 28
2.2.5 3G/​4G/​5G 28
2.2.6 LoRAWAN 29
2.2.7 Sigfox 29
vi Contents

2.3 Protocol Stack 29

2.3.1 CoAP 29
2.3.2 MQTT 31
2.3.3 AMQP 32
2.3.4 XMPP 33
2.3.5 Comparison of Application Layer
Protocols 33
2.4 Electronic Platforms 35
2.4.1 Arduino 35
2.4.2 Raspberry Pi 36
2.4.3 Beagle Bone 37

3 IoT–​Cloud Convergence 41
3.1 Introduction 41
3.1.1 Five Quintessential Characteristics of
Cloud Computing 42
3.1.2 Cloud Computing Service Models 43
3.1.3 Cloud Computing Deployment Models 43
3.2 Opportunities and Challenges 44
3.2.1 IoT Requirements to Meet the Future
Market Potential 44
3.2.2 How Cloud Comes in Handy? 46
3.2.3 Application Areas for IoT–​Cloud
Convergence 46
3.2.4 Challenges That Come with
Convergence 48
3.3 Architecture for Convergence 51
3.3.1 An Overview of Cloud-​based IoT
Services 51
3.3.2 State of the Art: Convergence
Architectures 55
3.3.3 A Simplified Convergence Model 58
3.4 Data Offloading and Computation 59
3.4.1 Data Offloading and Computation:
An IoT Perspective 59
3.4.2 Edge Computing Technologies for an
IoT Network Infrastructure 60
3.4.3 Data Offloading and Computations 63
3.4.4 Offloading Considerations and
Challenges 65
3.5 Dynamic Resource Provisioning 67
3.5.1 The Resource Provisioning Activity
and Requirements for IoT 67
3.5.2 Dynamic Resource Provisioning 70
 Contents vii

3.6 Security Aspects in IoT Cloud Convergence 72

3.6.1 ENISA’s Categorized Security
Challenges in IoT Cloud Convergence 72
3.6.2 IoT Security Designs Based on Edge
Computing 74
3.7 IoT–​Cloud Convergence: Test Beds and
Technologies 78
3.7.1 Overview of Testbeds and Platforms 78

4 Smart Computing over IoT–​Cloud 85

4.1 Introduction 85
4.2 Big Data Analytics and Cognitive Computing 85
4.2.1 Cognitive Computing Capabilities 86
4.2.2 Underlying Technologies 89
4.2.3 Empowering Analytics 94
4.3 Deep Learning Approaches 95
4.3.1 Artificial Neural Networks (ANN) 96
4.3.2 Convolution Neural Network (CNN) 97
4.3.3 Recurrent Neural Networks (RNN) 99
4.4 Algorithms, Methods, and Techniques 101
4.5 Case Studies 107
4.5.1 Health Care 107
4.5.2 Smart Home 109
4.5.3 Manufacturing 109
4.5.4 Retail Sector 110

5 Conclusion 113
5.1 Summary 113
5.2 Issues and Challenges 115
5.3 Future Outlook 116

Index 119
 A UT HO R S

Parikshit N. Mahalle obtained his B.E. degree in computer science

and engineering from Sant Gadge Baba Amravati University,
Amravati, India and M.E. degree in computer engineering from
Savitribai Phule Pune University, Pune, India. He completed his
Ph.D. in computer science and engineering with a specialization
in wireless communication from Aalborg University, Aalborg,
Denmark. He was a post-​ doctoral researcher at CMI, Aalborg
University, Copenhagen, Denmark. Currently, he is working as
professor and head in the Department of Computer Engineering
at Sinhgad Technical Education Society’s (STES’s) Smt. Kashibai
Navale College of Engineering, Pune, India. He has more than
20 years of teaching and research experience. He is serving as a
subject expert in computer engineering, Research and Recognition
Committee at several universities like SPPU (Pune) and SGBU
(Amravati).
He is a senior member of IEEE, ACM member, life member of
both CSI and ISTE. Also, he is a member in the editorial board of
journals, IEEE Transaction on Information Forensics and Security
and IEEE Internet of Things Journal. He is a reviewer for IGI
Global –​ International Journal of Rough Sets and Data Analysis
(IJRSDA), Associate Editor for IGI Global –​International Journal
of Synthetic Emotions (IJSE), and Interscience International
Journal of Grid and Utility Computing (IJGUC). He is a member of
the Editorial Review Board for IGI Global –​International Journal
of Ambient Computing and Intelligence (IJACI). He is also working
as an associate editor for IGI Global –​International Journal of
Synthetic Emotions (IJSE). He has also been a technical program
committee member for international conferences and symposium
x Authors

like IEEE ICC, IEEE INDICON, IEEE GCWSN, and IEEE

ICCUBEA.
He is a reviewer for the Springer journal Wireless Personal
Communications, reviewer for Elsevier journal Applied Computing
and Informatics, and member of the Editorial Review Board of the
journals IGI Global –​International Journal of Ambient Computing
and Intelligence (IJACI) and Global Research in Computer Science.
He has published more than 150 research publications having
1,372 citations and H index 16. He has been an author/​editor in five
edited books published by Springer and CRC Press. He has seven
patents to his credit. He has also delivered an invited talk on “Identity
Management in IoT” to Symantec Research Lab, Mountain View,
California. He has delivered more than 100 lectures at national
and international level on IoT, Big Data, and digitization. He has
authored 11 books on various subjects. Some of them are Context-​
Aware Pervasive Systems and Application (Springer Nature Press),
Design and Analysis of Algorithms (Cambridge University), Identity
Management for the Internet of Things (River Publications), Data
Structure and Algorithms (Cengage Publications), and Programming
Using Python (Tech-​Neo Publications MSBTE).
He was the Chairman of Board of Studies (Information
Technology), SPPU, Pune. He is now the Member –​Board of
Studies (Computer Engineering), SPPU, Pune. He has been a
member of the Board of Studies at several institutions like VIT
(Pune), Govt. College (Karad), Sandeep University (Nashik),
Vishwakarma University (Pune), and Dr. D. Y. Patil International
University (Pune). He has also been a technical program committee
member for many international conferences.
He is a recognized Ph.D. guide of SSPU, Pune, guiding 7
Ph.D. students in the area of IoT and machine learning. Recently,
two students have successfully defended their Ph.D. He is also
the recipient of “Best Faculty Award” by Sinhgad Institutes and
Cognizant Technologies Solutions. His recent research interests
include algorithms, Internet of Things, identity management, and
security. He has visited a few countries like Denmark, France,
Sweden, Germany, Austria, Norway, China, Switzerland, and
Singapore.

Nancy Ambritta P. graduated in computer science and engineering

from Anna University, Tamil Nadu, India, in 2010 and has her
master’s in computer engineering from Smt. Kashibai Navale
College of Engineering, Pune, India, in 2015. Her research interests
 Authors xi

are Cloud security and the Future Internet. She has about three
years of experience in the software engineering field and 3.5 years
of experience in teaching and research. She has published research
articles in international journals and conferences. She is a reviewer
for the Springer journal Wireless Personal Communications.

Gitanjali Rahul Shinde has overall 11 years of experience, presently

working as SPPU-​ approved Assistant Professor in Department
of Computer Engineering, Smt. Kashibai Navale College of
Engineering, Pune. She has done a Ph.D. in wireless communi-
cation from CMI, Aalborg University, Copenhagen, Denmark on
“Cluster Framework for Internet of People, Things and Services”
and awarded the Ph.D. on May 8, 2018. She had obtained M.E.
(computer engineering) degree from the University of Pune, Pune,
in 2012 and B.E. (computer engineering) from the University of
Pune, Pune, in 2006. She has received research funding for project
“Lightweight Group Authentication for IoT” carried out in SPPU,
Pune. She has presented a research article in the World Wireless
Research Forum (WWRF) meeting at Beijing. She has published
more than 40 papers in national and international conferences and
journals. She is author of three books and also editor of book The
Internet of Everything: Advances, Challenges and Application (De
Gruyter Press).

Arvind Vinayak Deshpande has obtained B.E. in electrical engin-

eering and Ph.D. in computer engineering from Shivaji University,
India. He is now the Principal of Smt. Kashibai Navale College of
Engineering since December 2003. He was selected at ISVOR-​FIAT
International Italy as a Master-​Trainer in 1998. He was selected as
chairperson to conduct a session on Image Compression in the inter-
national conference held at Marbella, Spain, in September 2004
by IASTED, Canada. He was a member of the Board of Studies
(BOS) of Computer Engineering, University of Pune, India. He has
signed a memorandum of understanding (MOU) with a Rwanda
nonprofit organization (NGO) for setting up an engineering college
and MBA institute at Kigali in Rwanda (Central Africa). He is chief
co-​coordinator of the Indian-​French student organization. He has
received several awards for his contribution to engineering edu-
cation at the state and national level, which has been recognized
by Indo-​American Society, Indian Society for Technical Education
(ISTE), IETE, and other voluntary organizations. He is a recognized
Ph.D. guide in computer engineering at SPPU.
 PR EFA CE

Failure will never overtake me when my definition to success is

strong enough!
Dr. A.P.J. Abdul Kalam

The IoT–​Cloud convergence has occupied the limelight in recent

years due to the emerging demands in IoT applications. The IoT
applications have evolved from merely simple standalone applications
such as home automation, process automation, or care for elders to
advanced ones that include oil and natural gas rig safety, IoT for agri-
culture, and many more. The resource requirements for processing
standalone applications can be sufficiently met by the participating
devices themselves. However, advanced IoT applications have
increased demand for resources due to huge volumes of data coming
in from various sources and the need for huge computations. This
cannot be met solely by the participating devices, thereby leading to
the convergence of Cloud into the applications.
The use of converged technology is growing enormously as there is
an enhancement in smart devices, communication, Internet of Things
(IoT), and Cloud computing. The convergence of IoT and Cloud com-
puting is required for various purposes in real-​time scenarios. The
fundamental requirement of IoT is resource sharing. Cloud com-
puting not only works upon the principle of sharing resources but
also emphasizes maximizing the availability of resources. One of
the key features of Cloud is the virtualization of underlying physical
devices, which makes it easier for users to share the same hardware in
a seamless manner. IoT has many mobile users and devices. Services
 P r e f ac e xiii

should be made available to these devices/​users independent of the

location. Cloud being location independent allows services to be
accessed from anywhere and from any device over the internet. The
mentioned advantages assure that the convergence of IoT and Cloud
can lead to many successful developments in the field. Considering
the above-​mentioned information, it is rightful to provide interested
readers with a book that efficiently presents the various technologies
and methods in an organized manner, enabling users to move grad-
ually from the basic to advanced concepts. This will enable novice
readers and researchers find all necessary information related to
IoT application development, which can lead to many productive
developments and findings in the area.
This book is a one-​stop shop that offers the readers everything
they need to know or use in a real-​time IoT application development
activity. The book offers a basic understanding of the IoT architec-
ture, use cases, smart computing, and the associated challenges in
designing and developing the IoT system. This provides a strong
foundation to a novice reader. All the technical details that include
information about the protocol stack, technologies, and platforms
used for the implementation are explained in a clear and organized
format. Various aspects pertaining to the IoT–​ Cloud computing
platform along with an introduction to the offloading and resource
sharing mechanisms are discussed in the book, which helps the reader
to gain an in-​depth understanding about the internal working of the
applications/​systems. Techniques and case studies that include smart
computing on the IoT Cloud models along with test beds for experi-
mentation purposes are introduced to the reader. An outlook from
where the readers can build upon and work toward developing their
applications is presented in the book.
This book provides challenges involved in IoT application and
discusses various technologies, protocols, and platforms that are
used in IoT application development. This book also summarizes
various testbeds that is a part of the FIESTA-​IoT project that aids in
experimenting new approaches toward the development and deploy-
ment of IoT applications. A few key features of this book are as
follows:

• Discusses the broad background of IoT and its fundamentals

• Discusses the broad background of IoT–​ Cloud conver-
gence architectures and its fundamentals along with resource
provisioning mechanisms
xiv P r e f ac e

• Emphasizes the use of context in developing context aware IoT

solutions
• Discusses various technologies, methodologies, and approaches
that play a prominent role in smart computing on the IoT Cloud
architecture
• Presents case studies of a few application areas of smart com-
puting over IoT Cloud
• Presents a novel C-​model that explains the IoT application
development phases in a simple yet effective way
• Presents a simplified convergence model that depicts the role of
Cloud in an IoT application.

In a nutshell, this book presents all information (basic and advanced)

that a novice and advanced reader needs to know regarding IoT appli-
cation development. The book also recommends the use of appro-
priate technology for better development of applications. The book
also contributes to social responsibilities by laying down the founda-
tion for the development of applications that can help in making day-​
to-​day activities easier by meeting the requirements of various sectors
such as education, healthcare, agriculture, and other vital aspects of
human lives.
newgenprepdf

A CKNO W LEDGM EN T S

We would like to thank many people who encouraged and helped us

in various ways throughout this book, namely our colleagues, friends,
and students. Special thanks to our family for their support and care.
We are thankful to the founder president of STES, Prof. M. N.
Navale, founder secretary of STES; Dr. (Mrs.) S. M. Navale, Vice
President (HR); Mr. Rohit M. Navale, Vice President (Admin);
Ms. Rachana M. Navale, our Principal; Dr. A. V. Deshpande, Vice
Principal; Dr. K. R. Borole; and Dr. K. N. Honwadkar for their con-
stant encouragement and great support.
We would like to thank the Prof. Dr. Nilanjan Dey for his con-
tinuous support and constructive guidance. We are also very much
thankful to all our department colleagues at SKNCOE, for their
continued support, help, and keeping us smiling all the time.
Last but not the least, our acknowledgments would remain incom-
plete if we do not thank the team of CRC Press, Taylor & Francis
Group, who supported us throughout the development of this book. It
has been a pleasure to work with the team and we extend our special
thanks to the entire team involved in the publication of this book.

Parikshit N. Mahalle
Nancy Ambritta P.
Gitanjali Rahul Shinde
Arvind Vinayak Deshpande
 1

INTR O D UCT I O N

1.1 OVERVIEW OF IOT

The term Internet of Things (IoT) was first introduced in 1999
by Kevin Ashton in the RFID Journal. Later, the International
Telecommunication Union (ITU) came up with a unique idea for con-
nectivity among the devices: Anyone will be able to have connect-
ivity with anything (device to device or human to device) at any time
or any place (with any other device or device in commute). In the
past, when this idea was discussed in the industry, the whole idea did
not seem realistic to them. However ITU and small business ventures
incorporated some sensors into the devices to make everyone see the
evolution of the internet along with the combination of technology
and vision. This new vision has the ability to change our day-​to-​day
life forever. This is one of the biggest reasons why IoT products and
their associated services will become popular in the society. IoT will
greatly impact every part of the life –​work, home, and social [1].
The definition of IoT is perceived differently, as many researchers
have different ideas about IoT. Because of the different definitions, it
has become truly difficult to comprehend the exact definition of IoT.
On the same lines, many scientists, researchers, and standardization
bodies are working on it but from their own perspective. This is creating
more perspectives in addition to what is present right now. A combin-
ation of these merged perspectives from the abovementioned bodies
can help us understand IoT and its ecosystem. IoT can be partitioned
into three different dimensions on the basis of multiple visions:

1. The first dimension consists of things which include key

chains, portable medical devices, or watches.
2. The second dimension is internet or connectivity-​
oriented
dimension. This includes communication and connectivity
among things.

1
2 Introduction

3. The third dimension is semantic. It means data collected by

various devices should be provided with some sort of tech-
nology in order to store, collect, sort and interpret that data.

The things can be further partitioned into different categories. Day-​

to-​day objects like books, wallets, and carts can be connected to
sensors, such as a temperature sensor and motion sensor. The others
can be combined with appliances like a refrigerator or air conditioner.
Sensors have been in use for a long time, but its connection with IoT
can be helpful. When the term IoT is referred, it means connectivity
among devices without human intervention. A vast amount of litera-
ture is available on IoT, but the difference in their perspectives causes
confusion.

1 . 1 . 1 Techn ical Buildin g Bloc ks

In the following section, the taxonomy of IoT components is
presented. There are three main components of IoT [2,3]:

1. Hardware, which consists of sensors and actuators

2. Middleware, which is provided for the purpose of data analysis
3. Presentation, which is visualization of information made easy
with the help of this layer.

Hardware, that is sensors and actuators, are the main components

of IoT. Sensors are used to collect information on parameters like
temperature, pressure, or humidity. Actuators are programmed to
take action based on the value sensed by the sensors. These sensors
can communicate through various wireless communication tech-
nologies like radio frequency identification (RFID), WiFi, and
Bluetooth. These sensors need to have unique identification to be
deployed in IoT application. WSN is the basic network in IoT. Data
sensed by these sensors are stored and processed using Cloud com-
puting. A few technical building blocks of IoT are discussed as
follows:

1.1.1.1 Radio Frequency Identification (RFID)

RFID is one of the technologies which can be embedded into the
microchip and can be used in wireless communication. Identification
of the things that are attached to it is a major task. There are two types
of RFIDs:
 Introduction 3

1. Active RFID: This RFID is powered. It means that it has

its own power source. It can also communicate if neces-
sary. Monitoring of valuable cargos is achieved by the active
RFID tags.
2. Passive RFID: This RFID does not use battery. It makes use
of readers’ signal to identify the ID for the RFID reader. The
application of this RFID can be seen in retail, transportation,
access control etc. These tags are used in various applications
like road toll, banking cards etc.

1.1.1.2 Wireless Sensor Networks (WSN)

WSN consists of a number of intelligent sensors. These sensors con-
tinuously gather information after which processing is done on the
collected data. The pre-​processed data is used for analysis so that
valuable information can be extracted from the collected data. Active
RFID and lower end WSN nodes serve the same purpose in terms
of processing capability and storage. The challenges that come with
WSN are multidisciplinary in nature and so is the potential of this kind
of network. Sensor data that is gathered from sensors is sent to the
centralized or distributed systems for analysis. Recent advancements
in technology have made this wireless communication very easy.

1.1.1.3 Addressing Schemes

Any IoT system should be able to uniquely identify things. This step
is very crucial in making that IoT system a success. This will help in
identification of billions of devices uniquely. Also, the management
of these devices with the help of internet will be possible. A major
critical aspect of creating a unique address is as follows:

• Uniqueness of address
• Reliability
• Persistence
• Scalability

Every “thing” that is connected in the system and those that are going
to join should have a unique address, location, identity, and func-
tionality. Currently the IPV4 addressing scheme is being used, but
as the number of the devices increases, it becomes impossible for
IPV4 to give unique ID’s to each and every device. The solution to
this problem is the IPV6 addressing scheme. However, this will only
solve the problem to some extent. The heterogeneity of the devices,
4 Introduction

their associated data types, and the operations on them still remains
a major concern. In order to channel the data traffic ubiquitously and
relentlessly, persistent network functioning can be used. Even if the
TCP/​IP is taking care of the efficient delivery from source to destin-
ation, IoT is still facing a bottleneck at the interface of the gateway
and the sensors. Scalability of the available address should be sustain-
able. If at all another network is required to join in, then the perform-
ance of the existing network should not be hampered in any way. To
solve these problems, Uniform Resource Name (URN) system was
developed, which is being considered as the base of the development
in IoT. This system creates replicas or mirror images of the devices
that is more easily accessible through the URL. As large volumes of
data are being created, the metadata can help in the transfer of this
data. IPV6 is also a good option to be able to uniquely identify things
and monitor and control them. Another important task is the develop-
ment of lightweight IPV6 for home appliances.

1.1.1.4 Data Storage and Analytics

The unprecedented outcome of any emerging field is the generation of
a large amount of data. The storage and management of the data, own-
ership, and expired data issues have gained importance. According
to statistics 5 percent of the totally generated energy is being used
by the internet. As days go by, this statistics will change; it simply
means the demand will increase. Hence the harvested energy using
data centers needs to be figured out plans for conserving energy. The
storage and use of data should be done smartly. It is imminent to
design AI algorithms that could be implemented in a centralized or a
distributed way. Fusion of algorithms needs to be designed to inter-
pret the collected data. In the state-​of-​the-​art various methods like
nonlinear, genetic algorithms, temporal machine learning methods
based on evolutionary algorithms, neural networks, and other artifi-
cial intelligence techniques are required to gain control of automated
decision making. These systems have abilities like interoperability,
integration, and adaptive communications. They also have modular
architecture in hardware as well as in software parts, which is required
for any IoT application.

1.1.1.5 Cloud Processing

The data is useless if there is no provision to store, collect, or analyze
that data. This is made possible with the help of Cloud computing.
The collected data needs to be analyzed to deduce if any action is
necessary, depending on the results. For example, when a person
 Introduction 5

is reaching home from work, then the thermostat can be turned on

depending on the distance between his current location and home, so
that when that person reaches home the temperature will be perfect.
On reaching the front gate, the gates will be opened and the person
can drive in. This is made possible because things collaborate and
take actions if necessary, which is made possible due to Cloud pro-
cessing. The thermostat and the front gate have sensors which sense
the current situation and act on it with the help of Cloud.

1.1.1.6 Security
Security is very important in any domain. In IoT the sensors, their
connecting interfaces, communication among them, infrastructure for
the system, Cloud processing, and storage require security at every
step. When IoT products were being developed, security was not the
biggest concern at the time. Because of this flaw many companies
still haven’t implemented the IoT measures. As time progressed, it
became evident that security in the system needs to be strengthened.
As hackers are tampering with vehicle control systems, for example,
more stringent security measures are needed to be implemented.
Hence many companies are asking for IoT products that have strong
security systems that will be difficult to breach.

1.1.1.7 Visualization
Visualization is a very important milestone in the creation of an IoT
application. This visualization helps users to easily communicate
with the environment. The displays (phones, tablets) nowadays are
designed to be more intuitive so that users will be able to reap full
benefit from it. For a layman, it is very important to get the infor-
mation needed in a more precise, compact, and easily understand-
able format. That is why visualization is very critical. As displays
are evolving from 2D to 3D versions, more information can be made
available to the user. This will in turn help policymakers to transform
data into meaningful knowledge. This consists of visualization and
detection of the associated raw data.

1 . 1 . 2 Io T i n Busin ess
IoT can change the perspective of business approach by providing IoT
connectivity. IoT will help businesses track and effectively monitor
assets in real time with the help of sensors [4]. Various industries
are shifting from manual working toward IoT due to automation and
smart applications.
6 Introduction

1.1.2.1 Intelligent Transportation

Transportation is one of the important domains where IoT can make
a difference. Using sensors, information about vehicles such as their
location, route, their physical condition, and many more can be
retrieved faster and without human intervention. This information can
be of help for safe and smart transportation. Further IoT can be used
in car networks to share high-​rate multimedia data. This type of net-
work is called as vehicular ad hoc networks (VANETs). Device-​to-​
device communication (D2D) is the pillar of such networks. In D2D
communication, devices will be able to communicate with very less
involvement of the network. This solves the latency issues to some
extent where vehicles will be able to communicate with each other
directly using sensors.

1.1.2.2 Smart Homes/​Buildings and Monitoring

Sensors come in handy for the prevention of hazardous situations.
The sensors can be installed at home to monitor the situation. For
example, environmental sensors can be used to check the quality of
air in the house such as levels of carbon monoxide concentration,
colour, gas leaks, humidity, hydrogen sulfide levels, and tempera-
ture remotely. Home automation is the best example of IoT appli-
cation in the households. With the help of IoT lights around the
house can be controlled. Also, if there are elderly people at home
then incorporation of IoT in day-​to-​day life can make life better.
Continuous monitoring of elderly people will become very efficient
with IoT.
To provide full safety to buildings, IoT can be incorporated in the
buildings. Some of the startups are developing sensors that can be
embedded into the foundation of the building to monitor the consist-
ency of load bearing, heating systems, and lift.

1.1.2.3 Smart Grids

In early days when the smart grids were installed, the main task of
the grid was to control supply and if necessary cutoff the supply of
power as per instructions. But incorporating the IoT in the grid can
provide the grid with the ability to execute a wide array of functions.
These grids will be able to remotely monitor not only the power
requirements but also the gas or water requirements. In future smart
grids will be able to replace sensors, which will be able to reduce cost
by relaying voltage and current measurements.
 Introduction 7

1.1.2.4 Environment Observation and Forecasting

Growing urbanization and harmful human activities are proving to be
detrimental to nature. To preserve the environment and all the life it
holds, IoT can be very helpful. There are some species that are on the
verge of extinction. Tracking of such species can be done with global
positioning system (GPS)-​enabled sensors. Preserving such species
will be a little bit easy with the sensors. Long-​range radio transmitters
can be used to monitor the ground habitats of the species. Because of
the extinction of some of the species, they have been termed as exotic.
Trespassing has also increased on the habitats for the possession of
such animals. IoT can help to ensure the safety of such animals.
Deforestation is also a grave threat to animal life. Because of
wildfires or human mistakes, it is rapidly increasing. IoT sensors can
be deployed in the dense forests, which can alert the system just in
time if any natural tragedy is about to occur. IoT can also help pre-
dict earthquakes or avalanche, so that the population at risk can be
removed from the place as soon as possible.

1.1.2.5 Smart Agriculture and Farming

Water is one of the most important needs of life. It has to be conserved
because reservoirs are depleting at faster rate. According to statistics
70 percent of the fresh water is required for farming purposes, but
60 percent of that goes to waste because of the failure in irrigation
system and faulty agricultural techniques. Sensors can be installed in
the irrigation system to detect the failures in the system. In this way,
whenever there is a failure in the irrigation system, immediate action
can be taken to stop the water wastage. IoT systems can be developed
for vineyards to sense temperature, soil moisture, and solar radiation
and alerts can be sent to the farmers in case of a dangerous situation.

1.1.2.6 Health Care

There are numerous fields available in which IoT can be applied. IoT
can lend a helping hand in the revolutionization of the health care
industry. If a patient needs continuous monitoring then the common
practice in health care industry is that a physician is required to
continuously monitor the patient. But recent trends in WSN along
with embedded systems, small monitoring device or sensors can be
implanted in the patient’s body. These sensors can form a network,
which is termed as the body sensor network (BSN).These networks
are very helpful in continuous monitoring of patients. Cardiac disease
8 Introduction

can be easily diagnosed if the patient’s ECG is monitored continu-

ously. This is one of the common applications of BSN. For other crit-
ical diseases also, BSNs can be used.

1.1.2.7 Education
At present, internet access is easily available to a majority of the popu-
lation around the world. Students and academicians around the world
can now easily access each other’s work. Internet has provided the
world with a wide platform on which experts can deliver the sessions
and brainstorming of ideas has become possible and fruitful. Beyond
that, experts are asked to deliver lessons on the respective fields
they are working on. With massive open online courses (MOOCs),
majority of the population is able to receive online education. The
idea of “flipped classrooms” is gaining popularity. In this, students
will learn the concepts outside the classroom so that during the
physical lecture innovative ideas can be discussed. The MOOCs are
turning out to be a saviors for people in developing countries where
many cannot afford the fees for higher education.

1.1.2.8 Smart Clothing

The wearable IoT is the next phase of IoT. The clothes will be
designed for a special purpose with sensors embedded in them. For
this, low-​volume fabrics are preferred. However, in future conductive
materials can be used in the production of cloth material. In order to
make the clothes responsive, placement of sensors is very vital. The
next important part is communication among the sensors. Lastly an
interface needs to be designed in order to collect, store, and interpret
the collected data. The tensile strength and the appearance of the cloth
can be determined with the help of the collected information.
In recent years, various IoT projects have been developed for the
benefit of society. A few prominent projects are discussed below:

1.1.2.9 Internet of Things-​Architecture (IoT-​A)

IoT-​A is a European project aiming to develop a reference archi-
tecture model for the IoT with interconnected technologies, smart
devices, resources, and services. IoT-​A is a group of university and
industry-​funded institutions and experts in September 2010 by the
European Union under the FP7 framework [5]. “It can be seen as
an umbrella term for interconnected technologies, devices, objects,
and services.” The main challenge of IoT-​A is to find a way to
connect different devices, applications, and technologies in a single
 Introduction 9

coherent network. To achieve this IoT-​A provides a set of common

protocols and interfaces and also the architecture to handle interoper-
ability, heterogeneity, and scalability issues of IoT. IoT-​A has three
models: (i) domain model, (ii) information model, and (iii) com-
munication model. Domain models define the IoT object attributes,
that is, name and identity and define the relation between devices.
Information model defines the methods to store and retrieve infor-
mation and communication model defines the protocol stack for
the IoT communication. Kramp and his team have developed IBM
infrastructure platform for wireless sensor networks and published
specifications for an embedded virtual-​machine byte code, platform
application programming interfaces (APIs), and protocols.
IoT-​A follows the four-​step process depicted in Figure 1.1. The pro-
cess discusses the underlying assumptions, motivation, and how the
existing IoT architecture can be used as a methodology. In the second
step, IoT-​A analyzes different business scenarios and stakeholders
to understand the requirements of the IoT-​A. A stakeholder analysis
helps to understand the aspects of the architecture model needed for
different stakeholders and their concerns. With the interoperability
support from multiple organizations and IoT reference models,
the new IoT-​A reference model is proposed. IoT-​A is the abstrac-
tion layer and understands communication between heterogeneous

Figure 1.1 IoT: A reference model.

10 Introduction

devices. IoT-​A reference architecture focuses on abstraction sets of

mechanisms other than application architecture.

1.1.2.10 Coordination and Support Action for Global

RFID-​Related Activities and Standardization (CASAGRAS)
Coordination and support action for global RFID-​related activities
and standardizations (CASAGRAS) was a project financed by the
EU focused on foundational studies on international questions about
RFID in support of IoT. It provided regulations and standardization
of RFID [6]. CASAGRAS defines the components of IoT as a global
network infrastructure, which is similar to the world-​wide internet
and allows messaging from one device to another. CASAGRAS has
focused on the need for identification of each physical object in the
IoT. It defines the virtual object, sensors, actuators, and connectivity.
The virtual object is a representation of a proxy relationship with the
physical objects. Sensors are devices that measure physical, chemical,
and other forms of quantitative data. Actuators are used to control the
system and are coupled with sensors. Connectivity is the interface
between source data and a device. CASAGRAS architecture model
depicted in Figure 1.2, consists of three layers as follows:

1. Physical layer
In the physical layer, object and things are identified. The
physical layer contributes functional components to the IoT,
that is, technologies that connect objects and carry data like
RFID, Zigbee, and Bluetooth.
2. Gateway layer
In the gateway layer, there are layers that act as a mediator
between connected devices and host management systems.
3. Application layer
Application layer sends user information or functional data
to the connected devices through gateway layer to support
applications and services.

1.1.2.11 MAGNET and MAGNET Beyond

My personal Adaptive Global NET (MAGNET) was an EU-​funded
project in the area of personal networks. MAGNET has developed
user-​centric system and business model concepts for secure person net-
work in the multi-​network, multi-​device and multi-​user environment
 Introduction 11

Figure 1.2 CASAGRAS architecture model.

[7]. MAGNET has 30 partners from 15 countries, which are highly

acknowledged industrial partners, universities, and research centers.
The objectives of MAGNET are to design and implement tech-
nologies and protocols that are needed to build a personal network
according to the user’s requirements in terms of quality and secure
service. It implements a multilayer security concept to ensure the
security and privacy of user’s data. It designs efficient, flexible, and
scalable air interface for the components of personal networks. It
collaborates with other projects like MATRACE, 4MORE, PHONIX,
WINNER, PULSERS, E2R, and AMBIENT NETWORK to imple-
ment user profiles, context and service discovery, PN to PN commu-
nication, trust and identity management, universal convergence layer,
PN federation, and commercialization of PN architecture.
A personal network is a concept of supporting professional and
private activities in an ad hoc network and infrastructure-​ based
dynamic networks. Users can access personal devices at anytime from
12 Introduction

Figure 1.3 MAGNET architecture.

anywhere as depicted in the Figure 1.3. Configuration and connect-

ivity of personal devices depends upon time and place and requires
resources. MAGNET Beyond takes the PN concept further to form
the PN federation (PN-​F). Devices of two different PNs communicate
with each other seamlessly and for a limited time to complete a spe-
cific task. The features of MAGNET features are listed below:

• MAGNET protocol is designed based on the Linux platform

and IPV4 networking.
• Networking is built using universe convergence layer. The
gateway node functionality was divided into two parts: the
gateway and edge router.
• Security for MAGNET PN prototype is based on certified PN
formation protocol and data is protected by encryption based on
keys derived using the certificates.
• Service discovery is done by the service management platform,
which provides context-​aware service discovery for personal
networks. Modified Universal Plug and Play (UPnP) device
modules were responsible for discovery, registration, and
control.
• MAGENT supports context management via secure context
management framework. SCMF is responsible for storing, pro-
cessing, and delivering context information.
 Introduction 13

Two applications developed to test MAGNET’s real use are Icebreaker

and Lifestyle Companion. The aim of the Icebreaker application was
to demonstrate personal network capabilities at a big event like a con-
ference. The user can utilize “Icebreaker service” offered by the con-
ference organizers. Users can access user services to subscribe and
check into the conference. The services also use the name, contact
information, and business interest to access equipment in a showroom
and to display presentations. The Lifestyle Companion application is
an electronic personal trainer. When the user enters the gym, it asks
the user to step onto a scale. The weight is taken and accordingly cer-
tain exercises are recommended.

1.1.2.12 BUTLER
BUTLER is the EU project with the aim of providing a set of
guidelines and principles to build IoT systems [8]. BUTLER effect-
ively performs integration of platforms and is extensible to add new
functional components to the architecture. BUTLER facilitates to
add value-​added services such as localization, context capturing,
and security management to build effective context-​ aware infor-
mation systems, which seamlessly operate across various scenarios
in the smart life environment. The major challenges are to provide
privacy and security in heterogeneous systems. It provides a service-​
oriented architecture for dynamic, interoperability, and complex event
processing.
BUTLER systems are deployed using three entities: smartObject,
smartServer, and smartMobile. BUTLER architecture has identified
four layers for all functionalities to fulfill requirements such as con-
text management, communication, services and device management,
depicted in Figure 1.4.

• The communication layer defines the end-​to-​end infrastructure

that connects smartObject, smartMobile, and smartServers.
• Data/​context management layer defines data models, interfaces,
and procedure for data collection from data sources. It also
helps to process the data in order to provide context informa-
tion and transfer the raw data to high-​level information.
• Device management layer defines the necessary components of
management, maintenance of smartObject, services, configur-
ation, software management, and diagnostics.
• Service layer defines interfaces for description, discovery,
binding, deployment, and context-​aware services.
14 Introduction

Figure 1.4 BUTLER context model.

BUTLER focuses on five scenarios as follows:

• Smart homes and offices

• Smart shopping
• Smart mobility and transport
• Smart healthcare and wellness; and
• Smart cities

1.1.2.13 Smarter Cities Data Management (SMARTIE)

SMARTIE is the EU project that aims to develop a framework for
IoT to store, process, and distribute the big data for use in smart city
applications [9]. SMARTIE framework provides security, privacy,
and trust by embedding security mechanisms in the IoT framework
instead of providing them as extra plug-​ins. It is based on IoT-​A archi-
tecture. SMARTIE components are organized into functional groups
like IoT-​A project. It has seven functional groups: application, man-
agement, service organization, virtual entity, IoT service, communi-
cation, and security functional groups. SMARTIE uses the Discovery
framework for discovery of the IoT resources using Constrained
Application Protocol (CoAP) and CoAP resource directory. Elliptic
curve cryptography (ECC) based distributed capability access con-
trol (DcapBAC) mechanism is used for access control. DcapBAC
 Introduction 15

uses XACML and JSON for lightweight approach, as IoT devices

are resource constrained. SMARTIE project is demonstrated for few
scenarios such as energy management, public transport, traffic man-
agement, and city information center.

1.2 IOT ARCHITECTURE: LAYERED PERSPECTIVE

IoT majorly consists of resource constraint devices. These devices
have different communication, computational, and storage capabil-
ities. The organization of these devices plays an important role in any
IoT application. IoT architecture follows the layered approach [10]
and 10C model, which is discussed in the next chapter. This layered
approach has four layers, as depicted in Figure 1.5. In this approach,
devices are at the bottom level, followed by the interconnection, ser-
vice, and application layers in order.

• Device Layer: This layer consists of heterogeneous devices.

These are heterogeneous in all sense, that is, size, communication

Figure 1.5 IoT layered architecture.

16 I n t r o d u c ti o n

capability, energy source, computational capacity, identifica-

tion, etc. In any IoT application, different devices may require
to work together. Due to this heterogeneity, organizing and
working of IoT application requires gateways that can com-
municate with all types of devices. Devices can communicate
with the help of Wi-​Fi, Zigbee, BLE, local area network (LAN),
Long Range Wide Area Network (LoRaWAN), and many more.
• Communication Layer: This layer provides communication
between users and IoT services provided by IoT devices. Users
can access these services using the internet, hence addressing
and identification of IoT devices is the basic challenge of this
layer. Various protocols can be used in this layer, that is, IPv6
Routing Protocol (RPL), 6LowPAN, IPv4/​IPv6, etc.
• Service and Cloud Layer: IoT services are developed based on
information sensed by IoT devices. The information storage,
information processing, and forming services are important
challenges of this layer. In various IoT applications, Cloud com-
puting is used for information storage and plays an important
role in service utilization. Users should be made aware of these
IoT services. Hence service broadcasting is also one of the
important functionalities of this layer. To enable this various
technologies can be used, that is, multicast domain name
system (mDSN), physical web, UPnP, DNS service discovery
(DNS-​SD), HyperCat etc.
• Application/​Authorization Layer: The functionality of this layer
is to provide IoT services to the users. Various application layer
protocols can be used, that is, MQTT For Sensor Networks
(MQTT-​ SN), Extensible Messaging and Presence Protocol
(XMPP), Constrained Application Protocol (CoAP), Message
Queuing Telemetry Transport (MQTT), Data-​ Distribution
Service for Real-​Time Systems (DDS), and Advanced Message
Queuing Protocol (AMQP). This layer is also responsible
for providing these services to only legitimate users. Service
owners have a set of access policies and authentication/​author-
ization protocols to validate users.

1.3 SMART COMPUTING

Smart computing plays a crucial role in IoT. In IoT devices are
interconnected through internet and can be automated or controlled
remotely. Further, in smart computing, IoT devices can take action
 Introduction 17

based on information gathered by the IoT network without human

intervention as in many applications humans are unreachable. These
actions can be taken based on learning techniques as traditional pro-
gramming may be a very expensive solution. IoT with smart com-
puting can enhance usefulness of IoT applications and can also
increase revenue.

Functionalities of IoT with Smart Computing

• Smart Home: IoT devices in refrigerators can estimate quan-

tities and expiry of food items on the basis of which notifications
can be sent for the purchase of the items. It also recommends
recipes based on the available food items. Sensors at the garden
measures humidity and controls irrigation accordingly. Based
on temperature inside the house the temperature of the air con-
ditioner is managed.
• Manufacturing Industries: Classification of faults can be done
based on sensor’s diagnostic data. Further prediction of device
failure can be done, which can be of help for device mainten-
ance. Various mechanisms can be suggested for minimizing the
energy consumption of battery-​operated devices based on the
energy consumption pattern.
• Intelligent Transport: The recommendation of the number of
buses required for a particular route can be done based on the
pattern of the crowd in public buses. Maintenance of vehicles
can be scheduled based on sensor readings. Traffic signal
timings can be changed dynamically based on the number of
vehicles at a particular side.
• Healthcare: Alarms can be raised if there is any change in the
regular pattern of a patient’s body sensor data. Early detection
of various diseases can be done by data analysis.

1.4 IOT DESIGN: ISSUES AND CHALLENGES

1 . 4 . 1 Desi gn Issue s
IoT applications are implemented based on the information provided
by the IoT sensors. These data may be very large (Big Data) or small
and may be kept on local servers or on the Cloud. These data are hetero-
geneous and can lead to various design issues for developing the IoT
applications [11–​16]. Few design issues of data are discussed below:
18 Introduction

• Volume: Volume is defined as the amount of data that is being

created and stored. It is increasing exponentially from terabytes
to petabytes as days are passing. As the size is increasing,
the storage capabilities are also making advancements. What
cannot be captured in earlier days is possible now. This big data
can be classified on the basis of the generated data (type of data)
and time at which it is time generated. Type of data is often
referred as the variety of data. Text data and audio data require
different technologies for processing.
• Velocity: The rate at which data is being created is called as the
velocity of the data. In legacy systems traditional data analytics
was done on the updates received on a daily, weekly, or monthly
basis. As real-​time data is being generated in larger volumes,
analyzing and processing this data is highly necessary in order to
make informed decisions. Time is a very important factor. Some
of the domains like retail, telecommunications, and finance are
able to create high-​frequency data. Data generated data from
the mobile app can be useful in offering more personalized ser-
vices to the user on the basis of user’s preference.
• Variety: Variety refers to the heterogeneous nature of the data
that is being created and stored. This data is categorized into
three main sections:
• Structured data
• Semistructured data
• Unstructured data
If organization of the data can be done with the help of existing
data models, then that data is called as structured data. The data
that is represented with the help of rows and columns are excel-
lent examples of structured data. Only 5 percent of the data is
available in such a format. If data cannot be organized using any
data model, it is known as unstructured data. Examples of such
data include video, text, and audio data. Semistructured data
lies in between structured and non-​structured data. Extensible
Markup Language (XML) is one of the best examples of
semistructured data.
• Veracity: Unreliability associated with data sources is referred
to as veracity. For example, sentiment analysis with the help
of social media platforms is bound to create uncertainty in the
results. Hence there is a need of a mechanism that will be used
to manage the uncertain and unreliable nature of data.
 Introduction 19

• Variability: Inconsistency in velocity of the big data causes flux

in the flow rate of data, which is called as the variability of the
data. Data is being created from a number of sources and man-
agement of such diversified data is really a big task. As data
is being generated from diversified resources, it has different
underlying semantics.
• Low-​value density: Raw data is not usable. Data needs to be
processed and analyzed in order to make sense out of it. Logs
from general stores can be analyzed to make new formation of
products. It will in turn help in fetching more business and rev-
enue. It can also be used to analyze a customer’s behavior.

1 . 4 . 2 Cha lle n ges

IoT has numerous challenges such as standardization, communication
and connectivity interface, and heterogeneity [17–​25]. Some of the
challenges are addressed below.

• Mobility: An IoT system is a collection of different devices

working in collaboration with each other. Every device has a
goal associated with it. For example, sensing leak in irrigation
system or a failure in AC unit. Collaboration requires more
complex functionalities. The devices are categorized into two
parts. The first is static and second is mobile. Organization of
IoT devices will enable the system to be designed with complex
abilities. The static devices are more manageable to work with
as compared to mobile devices.
• Heterogeneity: Heterogeneity is the base of IoT devices. These
devices vary in size, computational power, their capabilities,
interface for communication, power source, and a whole bunch
of parameters. To design an IoT service while working with
such heterogeneity is a real challenge.
• Scalability: The use of IoT devices in our day-​to-​day life is
increasing with time. It will continue to grow exponentially
in the future. IoT applications’ basic requirement is to have a
number of heterogeneous devices (those can static or mobile).
According to the application requirement, the number of
devices involved can change. To support such a type of hetero-
geneity, IoT needs to provide scalability and support.
• Energy efficiency: IoT systems are comprised of a number
of devices. All the devices operate on some kind of a battery
20 Introduction

source. These devices require power in order to sense data,

transmit data, process of data, and to take action if necessary.
IoT systems should be designed to be more energy efficient so
as to sustain the life of devices.
• Security: The IoT systems are extremely vulnerable to attacks
because of the following reasons:
• Most of the devices are in the open and unguarded; hence it
is very easy to manipulate them.
• Wireless communication happens between these devices. It
is more vulnerable to attacks and very simple to compromise.
• As most of the devices are operated on a battery source,
complex techniques cannot be implemented even for
security purposes.
• Privacy and context management: The service providers at pre-
sent operate on “take it or leave it” strategy. It means if you
want a particular service, there are policies and conditions that
are designed by the service providers for sharing personal data.
If you want the service, you have to accept these terms. These
terms lead to the invasion of privacy for a typical user. Hence
sharing of the personal data with a number of service providers
will surely result in the violation of privacy for the user. Hence
there is a need of IoT applications to figure out what the user
wants with minimal amount of data shared by the user.
• Context management and knowledge retrieval: The basis of any
IoT application is contextual information. After acquiring the
information, knowledge retrieval is the next main task. But both
these tasks face challenges because of the heterogeneity of con-
textual information.

REFERENCES
1. Gubbi, J., Buyya, R., Marusic, S., and Palaniswami, M. (2013).
Internet of Things (IoT): A vision, architectural elements, and
future directions. Future Generation Computer Systems, 29(7),
1645–​1660.
2. Lee, I., and Lee, K. (2015). The Internet of Things (IoT):
Applications, investments, and challenges for enterprises. Business
Horizons, 58(4), 431–​440.
3. Serbanati, A., Medaglia, C. M., Ceipidor, U. B., and Turcu, C.
(2011). Building blocks of the Internet of Things: State of the art
 Introduction 21

and beyond. Deploying RFID-​Challenges, Solutions, and Open

Issues, 251–​366.
4. Shinde, G.R. (2017). Cluster Framework for Internet of People,
Things and Services.
5. S. Meissner, S. SAP, and T. (2012) NEC, “Internet-​of-​Things
Architecture IoT-​ A Project Deliverable D2. 5-​ Adaptive, fault-​
tolerant orchestration of distributed IoT service interactions”.
6. I. G. Smith, Anthony Furness, Ken Sakamura, Ricky Ma, Eldor
Walk, Craig Harmon, Patrick Guillemin, Yong-​Woon Kim, Paul
Chartier, David Armstrong, Y.-​W. Kim, P. Consultants –​Paul
Chartier, and R. Ltd -​David Armstrong, “2 CASAGRAS –​Final
Report The CASAGRAS partnership CASAGRAS (Grant
Agreement 216803) is a Coordination and Support Action for
Global RFID-​related Standardisation Activities involving, in par-
ticular, organisations from.” Eu Fp7 Project Casagras, “Casagras
Final Report: RFID and the Inclusive Model for the Internet of
Things,” 2009.
7. R. Prasad, Ed., My Personal Adaptive Global NET (MAGNET).
Dordrecht: Springer Netherlands, 2010.
8. “Project Deliverable Project Number: Project Acronym: Project
Title: Integrated Project Internet of Things Title D3.2 Integrated
System Architecture and Initial Pervasive BUTLER proof of con-
cept,” 2013.
9. P. Langendörfer and J.-​ M. Bohli, “At A Glance: SMARTIE
Secure and sMArter ciTIes data management Project Coordinator
SMARTIE: Secure and smarter cities data management,” 2013.
10. M. Westerlund, S. Leminen, and M. Rajahonka (2014). Designing
business models for the Internet of Things, Technology Innovation
and Management Review, 4(7), 5–​14.
11. Labrinidis, A., and Jagadish, H. V. (2012). Challenges and oppor-
tunities with big data. Proceedings of the VLDB Endowment,
5(12), 2032–​2033.
12. Sivarajah, U., Kamal, M. M., Irani, Z., and Weerakkody, V. (2017).
Critical analysis of Big Data challenges and analytical methods.
Journal of Business Research, 70, 263–​286.
13. Bhadani, A., Jothimani, D. (2016), Big data: Challenges,
opportunities and realities, In Singh, M.K., and Kumar, D.G.
(Eds.), Effective Big Data Management and Opportunities for
Implementation (pp. 1–​24), Pennsylvania, PA: IGI Global.
14. Nasser, T., and Tariq, R. S. (2015). Big data challenges. Journal of
Computer Engineering & Information Technology, 4, 3. doi: http://​
dx. doi. org/​10.4172/​2324, 9307(2).
22 Introduction

15. Jagadish, H. V., Gehrke, J., Labrinidis, A., Papakonstantinou,

Y., Patel, J. M., Ramakrishnan, R., and Shahabi, C. (2014). Big
data and its technical challenges. Communications of the ACM,
57(7), 86–​94.
16. Jaseena, K. U., and David, J. M. (2014). Issues, challenges, and
solutions: big data mining. CS & IT-​CSCP, 4(13), 131–​140.
17. Rose, K., Eldridge, S., and Chapin, L. (2015). The internet of
things: An overview understanding the issues and challenges of a
more connected world. The Internet Society (ISOC), 22.
18. Tamane, S., Solanki, V. K., and Dey, N. (Eds.). (2017). Privacy
and Security Policies in Big Data. Pennsylvania, PA: IGI Global.
19. Vimal, S., Khari, M., Crespo, R. G., Kalaivani, L., Dey, N., and
Kaliappan, M. (2020). Energy enhancement using Multiobjective
Ant colony optimization with Double Q learning algorithm for
IoT based cognitive radio networks. Computer Communications,
154, 481–​490.
20. Dey, N., Fong, S., Song, W., and Cho, K. (2017, August).
Forecasting energy consumption from smart home sensor network
by deep learning. In International Conference on Smart Trends
for Information Technology and Computer Communications (pp.
255–​265). Springer: Singapore.
21. Mukherjee, A., and Dey, N. (2019). Smart Computing with Open
Source Platforms. Boca Raton, FL: CRC Press.
22. Mukherjee, A., Panja, A. K., and Dey, N. (2020). A Beginner’s
Guide to Data Agglomeration and Intelligent Sensing.
London: Academic Press.
23. Mhetre, N. A., Deshpande, A. V., and Mahalle, P. N. (2016).
Trust management model based on fuzzy approach for ubiqui-
tous computing. International Journal of Ambient Computing and
Intelligence (IJACI), 7(2): 33–​46.
24. Babar, S., Mahalle, P., Stango, A., Prasad, N., and Prasad, R.
(2010, July). Proposed security model and threat taxonomy for the
Internet of Things (IoT). In International Conference on Network
Security and Applications (pp. 420–​429). Springer: Berlin.
25. Mahalle, P. N., Anggorojati, B., Prasad, N. R., and Prasad, R.
(2013). Identity authentication and capability based access con-
trol (iacac) for the internet of things. Journal of Cyber Security
and Mobility, 1(4): 309–​348.
 2

INT ER NET O F T HI N GS
APPLICATIO N DEVELO PM EN T

2.1 APPLICATION DEVELOPMENT PHASES

Due to rapid advancements in e-​commerce and m-​commerce as well
availability of internet at a faster and cheaper rate, the use of mobile
applications and IoT-​enabled applications is increasing at a quicker
rate. IoT solutions have become an integral part of everyday life and
IoT application development has vast becoming potential in business.
Smartphones, sensors, cameras, and RFID are the major components
in all IoT use cases. Minimum human intervention for exchange of
information and interconnecting devices using wired and wireless
communication medium are the key steps in building IoT applications.

2 . 1 . 1 Pri nciples
In this section, a brief description of the main principles to be taken
into consideration before developing IoT applications is given. These
principles cover issues like secure and real-​time data collection, data
management, IoT platforms, and IoT–​Cloud convergence and they
are discussed below:

1. Secure data collection: In many IoT applications, data

is collected from a very large number of sensors that are
deployed in a closed environment or also distributed over sev-
eral locations. In such cases, data collection in a secure way
becomes a major issue. Necessary mechanisms for ensuring
data integrity and lightweight encryption for sensitive data
must be adapted.

23
24 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t

2. Real-​time data: In all scalable IoT applications, data is

collected in real time for processing and therefore traditional
packet transfer mechanism for data transfer is inadequate. For
efficient streaming, efficient techniques must be adapted for
collecting real-​time data.
3. IoT platforms: Data received from IoT devices is quite huge
and it is collected from heterogeneous devices. To store and
process this data, an appropriate IoT platform must be selected
in advance.
4. Cloud-​centric development: Centralized storage and fast
delivery of responses to the user are main characteristics of
IoT applications. Cloud-​centric application development is
more appropriate in such cases, which also optimizes the cost
and is also suited for carrier networks.
5. Data management: Due to rapid advancement in semicon-
ductor technology, space is getting cheaper with time. The
main challenge is how to make sense out of this big data. Data
science and machine learning techniques must be adapted for
effective data management.

2 . 1 . 2 C Mode l
This section presents IoT application development phases using
a novel C model. With reference to the layered perspective of IoT
architecture presented in Chapter 1, there are mainly three layers
in IoT: device layer, access layer, and application layer. The 10 C
model proposed in this book will help the reader understand various
phases and operational steps to be taken care during IoT application
development.

(a) Category
Before we start developing any IoT application, it is
important to decide whether the underlying IoT application
is representing an indoor or outdoor use case, as there are
different design issues for these use cases and they also have
different sets of requirement and challenges.
(b) Component
The finalization of category and requirements of the use case
helps to decide the components that are required for develop-
ment. The major IoT components include smartphones, RFID,
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 25

sensors, cameras, and various platforms that are required to

process the captured data.
(c) Cost
Cost is an important factor from the business perspective and
therefore it is better to estimate the cost in the early phases of
development. Types of components, their configurations, and
number of units of each component help to estimate the cost.
(d) Connectivity
Establishing connectivity between different devices is the first
and important step in IoT application development. Devices
could be connected to the user using different communica-
tion technologies such as Ethernet, WiFi, Bluetooth, WLAN,
or Zigbee. These technologies differ in communication range,
energy requirement, data rates, and interface.
(e) Communication
Once the devices are connected, communication is initiated
between the devices. Communication can take place in mul-
tiple forms like between two devices, between the device and
user, between two users, and between the user and device.
(f) Cooperation
For enabling the operation of various IoT services, multiple
sensors are used for communication. This multi-​device com-
munication helps to initiate cooperation and is required for
service discovery, device discovery, and availing services
from them.
(g) Convergence
Devices can use multiple communication technologies,
resulting in convergence. For example, when a device is in an
indoor scenario, then it is connected to the internet using WiFi
and in outdoor use cases, 4G or 5G cellular network is used.
(h) Content
Every device connected to the internet generates some data
and posts it on the Cloud. There are many plug-​and-​play
devices that can download data from the Cloud. These data-​
centric operations are periodical or real time and are based on
the type of application.
(i) Context
Context plays an important role in ubiquitous applications.
The context of a use case can be decided based on the con-
tent generated from the application. For example, data from
26 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t

applications like health care, agriculture, and smart home can

be used to easily decide the context and related information.
(j) Complexity
There is complexity associated with IoT applications and it is
applicable to all aforementioned phases in application devel-
opment. Time, space, and skill requirement also decide the
complexity of the application.

2.2 WIRELESS TECHNOLOGIES FOR IOT

Wireless technologies play a key role in all IoT use cases. In indoor
use cases, all devices are connected to the gateway device using a var-
iety of options, which include Bluetooth, Zigbee, or WiFi. In outdoor
use cases, options are WiFi or 2.5G/​3G/​4G.

2 . 2 . 1 Blu etooth , Blue tooth Low En e r gy, an d

Blu eto o th 5
Bluetooth uses 2.4 GHz of operating frequency available in Industrial,
Scientific and Medical (ISM) band and is mainly used to initiate com-
munication between two devices. The operational range of Bluetooth
is 2,400 to 2,483.5 MHz globally working with 79 channels where
each channel has a range of 1 MHz. The maximum range offered by
Bluetooth is 10 meters and it uses a client–​server model to enable data
transfer in full duplex mode at the data rate of 64 kbps [1]. A Piconet
network of devices is used for Bluetooth communication supporting
seven devices where one is a master and others are slaves.
However, current consumption in Bluetooth is 30 mA and to
overcome this limitation, all latest devices are using Bluetooth Low
Energy (BLE) where the current consumption is 15 mA. The manu-
facturer interoperability, more range for communication, low-​cost
implementation, and lightweight suite are main advantages of BLE.
BLE is more suitable for industrial applications where the communi-
cation range is 250 meters. BLE is not backward compatible with the
devices supporting Bluetooth.
The emerging prominent standard of Bluetooth more suited for IoT
application development is Bluetooth 5. It is revised and customized
based on the design issues and requirements for IoT use cases offering
a data rate of 2 Mbps, which is twice that of BLE. The transmission
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 27

power level in Bluetooth 5 is enhanced to +20 dBm with increased

advertised capacity, which is eight times more than BLE, making it
more suitable to create interactive IoT applications.

2 . 2 . 2 Zi g b ee
Zigbee is built on the top of IEEE 802.15.4 and is mainly used in
the wireless personal area network (WPAN) [2]. It consists of two
key layers: (i) physical layer and (ii) medium access control sub-​
layer. Router, end device, and coordinator are three main actors
in Zigbee. The coordinator acts as a responsible device for data
transmission activities and storage. End devices are generally tiny
devices with resource constraints like limited power, memory, and
computational power and have the ability to communicate with the
parent device. Zigbee is suitable for indoor use cases like smart
home or smart shop for monitoring and controlling purposes.
Zigbee operates at frequencies of 868 MHz, 902–​928 MHz, and
2.4 GHz and the data rate for communication is 250 kbps. The typ-
ical communication range is 30 meters for indoor and 100 meters
line of site for outdoor applications. Each Zigbee network can have
64Knodes.

2 . 2 . 3 Wi Fi
WiFi is a widely used wireless communication technology used
in both indoor and outdoor use cases. It is based on 802.11, which
represents the widely known standards for wireless local area
networks. Access points or hotspots are used to connect devices to the
internet providing a data rate of 54 Mbps. The operating frequencies
used by WiFi is 2.4 and 5 GHz in ISM band and provides commu-
nication range of 20 meters and 100 meters for indoor and outdoor
use cases, respectively. It provides interoperability for various WiFi
devices, is cost effective, and integrated with all widely used devices
ranging from laptops to TV. WiFi functions in star topology where
internet gateway is the center node and consists of a physical layer and
medium access control layer [3]. The main specifications of 802.11
are 802.11a (operating frequency: 5 GHz, speed: 54 Mbps), 802.11b
(operating frequency: 2.4 GHz, speed: 11 Mbps), and 802.11g (oper-
ating frequency: 2.4 GHz, speed: 20 Mbps).
28 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t

2 . 2 . 4 6 L o wPA N
Due to the emergence of IoT-​enabled applications where resource
constraints are the main challenge, low-​power and lossy networks
become a good candidate. The objective of these networks is to save
energy by managing other constraints like memory and duty cycle.
IPv6 over low-​power wireless personal area network (6LoWPAN)
[4] is a key option to meet the requirements of resource constraints.
Energy saving can be carried out by optimizing the amount of data
transmission. In resource-​constrained networks like IoT, requirement
of data transmission is less than 1280 bytes, which is the transmission
rate of IPV6.These networks use IEEE 802.15.4 physical layer where
the rate is 178 bytes per frame. In view of this, 6LoWPAN header
compression is required using IPV6 packets over 802.15.4.

2 . 2 . 5 3 G/​4 G/​5 G
The internet is getting faster and cost is becoming cheaper gradually,
the credit for which goes to advancements in broadband technology.
3G/​4G/​5G represent the generations of wireless technology that are
related to the speed of internet and are dependent on the incoming
signal strength. This wireless communication technology is mainly
used in outdoor IoT use cases and 2G is the first generation in this
family, which is almost obsolete now. The detailed specifications of
these generations are as follows [5]:

3G: Introduced in 2001, this generation is built on universal

mobile telecommunication system as the core network back-
bone. It provides speed up to 14 Mbps and operating range is
2,100 MHz. It has broadband capabilities with a bandwidth
capacity of 15–​20 MHz.
4G: Introduced in 2009, it has an enhanced data rate that
differentiates 4G from 3G and is mainly based on technologies
like multiple input multiple output and orthogonal frequency
division multiplexing. The main 4G standard is WiMAX. 4G
supports a top speed up to 1 Gbps and an average speed up to 50
Mbps. Due to these enhancements, 4G can support multimedia,
voice, and interactive communication. 4G is mainly used in ad
hoc, multi-​hop, and scalable mobile networks.
5G: Introduced in 2018, 5G operates in the range of 30–​300
GHz with low latency and broader coverage due to small cells
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 29

operated in radio millimeter band. The internet service provided

by 5G is World Wide Web and supports high-​end applications
like real-​time streaming of videos.

2 . 2 . 6 L o RAWA N
Low-​ power wide area network (LPWAN) [6] uses LoRaWAN
protocol, which support requirements for IoT applications with
mobility, less power, and two-​way communication. An improved net-
work capacity and optimized energy consumption are key features of
LoRaWAN supporting a data rate of 0.3–​50 kbps. LoRaWAN operates
in the band of 2.4–​5 GHz and uses star topology. For IoT applications,
this protocol is also supported by three types of unique keys: network,
application, and device keys for security purposes.

2 . 2 . 7 S i g fox
Sigfox is a unique service provider solution for IoT applications
offering software-​based Cloud-​centric communication solutions with
optimized energy consumption and less cost. Lightweight protocol,
small payload, star topology, and ultra-​narrow band radio modulations
make this solution a better choice for IoT applications. It is also reported
that Sigfox is more scalable as compared to other solutions [7].

2.3 PROTOCOL STACK

Different working groups such as the Internet Engineering Task Force
(IETF), World Wide Web Consortium (W3C), Institute of Electrical
and Electronics Engineers (IEEE), EPCglobal, and the European
Telecommunications Standards Institute (ETSI) provide diversified
protocols for each layer of the IoT as depicted in Figure 2.1 such
as the device/​physical layer, communication layer, service layer, and
application layer. In this section, Application Layer protocols for IoT
are discussed as highlighted in Figure 2.1.

2 . 3 . 1 Co AP
Constrained application protocol (CoAP) is an application layer
protocol developed by IETF intended to be used in resource-​
constrained IoT devices [8]. It is designed for “machine to machine
communication” type of applications such as automated smart
30 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t

Figure 2.1 IoT protocols.

building and automated service center. CoAP runs on devices that

support User Datagram Protocol (UDP) communication. It is
designed for simplicity, low overhead, and multicast support in
resource-​constrained environments. CoAP is a lightweight protocol,
which is intended to be used over HTTP similar to the representa-
tional state transfer (REST). It provides simple request/​ response
interaction model between application devices. It uses the Uniform
Resource Identifier (URI) to identify a resource. CoAP has a two-​
layered architecture –​a bottom layer designed to deal with UDP data
communication and asynchronous message communication. This
layer uses back off exponential for reliable data communication and
also detects message duplication. The upper layer deals with request/​
response communication over HTTP. CoAP consists of four types of
messages –​non-​confirmable, confirmable, reset (RST), and acknow-
ledgment (ACK) –​and four types of working modes: confirmable,
non-​confirmable, piggyback, and separate mode as represented in
Figure 2.2. In the confirmable mode, the server sends ACK to the
client after receiving the message. In the non-​confirmable mode, ACK
is not sent from server side, that is, unreliable communication. In the
piggyback mode, ACK is sent with next message and in a separate
mode ACK is sent by the server as a separate packet and the server
needs to take fix time to send an ACK. Hence the client waits for ACK
before sending the next message. Security is provided using Datagram
Transport Layer Security (DTLS). There are three main components
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 31

Figure 2.2 CoAP modes: (i) confirmable, (ii) non-​confirmable, (iii)

piggyback, and (iv) separate.

while considering security for CoAP: integration, authentication, and

confidentiality. DTLS also avoids cryptographic overhead problems
that occur in low-​layer protocols. In the literature, CoAP is used in
many IoT applications [9][10].

2 . 3 . 2 MQTT
Message queuing telemetry transport (MQTT) is a publish/​subscribe
lightweight messaging protocol [11]. It is used in applications where
devices are power constrained and low network bandwidth is avail-
able. It is designed for unreliable networks and high-​latency devices.
It was designed to handle a large amount of remote measurement data
in low-​bandwidth networks. MQTT has three components: publisher,
subscriber, and a broker, as depicted in Figure 2.3. MQTT is a publish/​
subscribe model type of protocol, where every device, for example,
temperature sensor, humidity sensor, or the light sensor is a publisher.
32 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t

Figure 2.3 MQTT.

A publisher always establishes a network connection with a

broker. The publisher and subscriber communicate to each other over
TCP/​IP network. Publishers are connected to the subscriber through
a broker. Every publisher either publishes the message to the broker
or subscribes to the messages from the broker. The data carried over
the MQTT protocol is called Application message. The label attached
to the application message is called Topic. The broker receives all
messages from the publisher and then it filters out the messages and
sends them to the interested subscriber based on “topic” name. MQTT
can add additional security by adding encryption to the application
data. We can pass a user name and password to the MQTT protocol
for security. MQTT is used in many real-​time IoT applications such
as intelligent parking [12] and fire alarm [13].

2 . 3 . 3 AMQP
Advanced message queuing protocol (AMQP) is an open standard
message-​oriented, publish/​subscribe, reliable, and secure protocol for
“machine to machine” (M2M) communication. AMQP runs over the
transport layer protocol. AMQP works similar to the email server and
follows the store and forward architecture [14].
In AMQP, the publisher sends a message to the subscriber through
a broker. The broker is divided into two parts: exchange and message
queue, as depicted in Figure 2.4. The exchange has the same function
as a transfer agent in the email server and message queue as a mailbox.
The exchange is responsible for receiving published messages by
different publishers and to distribute the received published messages
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 33

Figure 2.4 AMQP.

to the respective queues based on roles and conditions of each sub-

scriber. The message queue stores the messages on a disk. The sub-
scriber accesses messages from the message queue.

2 . 3 . 4 X MPP
Extensible messaging and presence protocol (XMPP) is an open and
XML-​based protocol [15]. It was designed for chatting and sending
small instant messages that are delivered immediately to the user. It
uses both client/​server and publish/​subscribe architecture, wherein it
wholly depends upon the application to choose suitable architecture for
its application. It is designed for real-​time applications and it supports
small memory footprint as well as low-​latency message exchange. The
XMPP architecture is similar to an email server, where anyone can run
their own XMPP server enabling to take individual control. XMPP
is secure using simple authentication and security layer (SASL) for
authentication and transport layer security (TLS) for a secure channel.
It adds end-​to-​end encryption to raise the security bar even further.
XMPP runs over TCP/​IP network. An XML message can create a lot
of overhead to the network due to the headers, which increases power
consumption, making it unsuitable for IoT applications.

2 . 3 . 5 Co m parison of A pplic ation L ay er

Pro to co ls
In the literature, AMQP is compared with MQTT by Luzuriaga et al.
[16]. Use of AMQP protocol is recommended for reliable and scalable
 newgenrtpdf
34
Internet
of
T h i n g s A pp l ica t i o n D e v e l o p m e n t
TABLE 2.1
A Comparison of Application Layer Protocols
Protocol Restful Transport Security Request/​Response Architecture Communication
CoAP Yes UDP DTLS Yes Tree D-​ to-​ D
Yes TCP SSL No Tree D to D
D-​ to-​ C
AMQP Yes TCP SSL No Tree D-​ to-​ D
D-​ to-​ C
C-​ to-​ C
XMPP Yes TCP SSL Yes Client–​server D-​ to-​ D
D-​ to-​ C
D-​to-​D: Device to Device, D-​to-​C: Device to Cloud, C-​to-​C: Cloud to Cloud.
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 35

messaging over WLAN. In Table 2.1, the application layer protocols

are compared and evaluated based on the parameters like transport
layer protocol used, architecture followed, type of communication
supported, security protocol used, that is, Datagram Transport Layer
Security (DTLS) and Secure Sockets Layer (SSL). Different applica-
tion protocols may be suitable for different IoT scenarios. Hence, it is
not feasible to recommend a single protocol for all IoT applications.
In the research work, Representational State Transfer (RESTful)
approach is applied for usage of IoT services.

2.4 ELECTRONIC PLATFORMS

As mentioned earlier, every IoT use case collects data from multiple
sensors and there is a need for electronic platforms to process this
data before it is forwarded to the gateway device for further pro-
cessing. This section provides a brief description of different elec-
tronic platforms and also a comparison of these platforms [17].

2 . 4 . 1 Ard u in o
Arduino is an 8-​bit microcontroller board based on the ATmega3280P
chip from the Microchip family. Arduino is an open source device
platform used for building small electronics boards. Arduino
supports software programming languages like C, C++, Python,
and JavaScript. Arduino hardware board is available for a price of
$20. The typical Flash code size is 32 kb and the RAM size is 2 kb.
The power consumption of an Arduino is typically 20 mA. Arduino
operates on 16 MHz clock frequency. Arduino does not have enough
memory to build large embedded systems. Arduino boards are suit-
able for the projects where cost and power consumption are the pri-
mary factors. Different Arduino hardware boards are available with
varying numbers of General-​Purpose Input/​Output (GPIO) pins and
communication media such as Bluetooth, WiFi, and Controller Area
Network (CAN).
Arduino is used in small projects like automatic room temperature
control using temperature and humidity sensors, detecting motion,
and controlling an automated soldering robot. Arduino boards are
used to read various sensor data and controlling relay-​based action or
sending sensor data to the Cloud using the WiFi or Bluetooth module.
There are various applications that can be implemented on Arduino
boards. Typical IoT Applications on Arduino boards are automatic
36 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t

watering to home plants, door open and closing from internet,

home automation, and motor controlling applications. Arduino is a
microcontroller-​based platform that does not have an RTOS system
installed. Arduino is not suitable for large projects, where the real-​
time operating system is required, or large data processing and huge
data storage are required.

2 . 4 . 2 Ra sp berr y Pi
Raspberry Pi is a 32-​ bit microcontroller board based on the
ARM1176JZF-​S architecture. It consists of BCM2835 system on
chip (SOC) from the Broadcom family and it is a multimedia chip.
A low-​cost single-​board computer (SBC) board is used in Raspberry
Pi and it is mounted on a Linux operating system. It uses an open
source operating system called Raspbian, which runs a pool of open
source software and is available for a price of $35.
The typical onboard storage for Raspberry Pi board is SD, MMC,
and SDIO card slot and 256 Mb SDRAM memory. The power con-
sumption of the Raspberry Pi board is typically 700 mA. Raspberry Pi
operates on 700 MHz clock frequency. Raspberry Pi board does have
Ethernet, USB 2.0, High Definition Multimedia Interface (HDMI)
video output, audio output, and dual-​core multimedia coprocessor.
Raspberry Pi boards are tiny and portable in size, making them suit-
able for projects based on embedded architecture. They are also a
good option for projects where high interactivity is a priority. In the
academic field, Raspberry Pi is more popular for designing innova-
tive projects in learning, bringing novelty in content delivery, and
schooling projects. While Gbit Ethernet and WiFi connectivity
make the Pi an ideal solution in an industrial IoT environment, the
Raspberry Pi board is used in IoT-​based intensive care unit (ICU)
patient monitoring systems, automatic door control based on lighting,
industrial automation, motion video, and a Web server.
One of the limitations of Raspberry Pi models is that they don’t
have a built-​in real-​time clock. Raspberry Pi board lacks interfaces
including RS-​485/​232 serial ports, CAN, and economical 1-​wire buses,
and digital and analog inputs and outputs. The Raspberry Pi board
being used in harsh industrial environments with high temperatures
and strong vibrations leads to performance losses and damage to the
board. The microSD cards can only retain a limited amount of infor-
mation, making them unsuitable for industrial applications involving
the storage of large amounts of data.
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 37

2 . 4 . 3 Bea gle Bon e

BeagleBone Black is a 32-​bit microcontroller board and uses ARM
Cortes A8 as its key component in the architecture. It consists
of Sitara AM335x SOC processor from the Texas family. The key
characteristics of this board are less cost and they support develop-
ment platform, which is community-​supported. These boards are
useful for low-​power applications and supports open source software
development. BeagleBone Black board is available for a price of $50.
The typical on board storage for the BeagleBone Black board is 4-​
Gb E-​MMCS and 512 Mb DDR3 RAM memory. The power consump-
tion of the BeagleBone board is typically 500 mA. BeagleBone Black
operates on a 1 GHz clock frequency. BeagleBone black board does have
a 3D graphics accelerator, neon floating point accelerator, fast Ethernet,
USB host, micro-​HDMI video output, and audio output and PowerVR
SGX530 coprocessor. BeagleBone is preferred by beginners due to its
simplicity, size, and is more suited for designing and implementing
IoT applications. Other key characteristics of BeagleBone compared
to Raspberry Pi are simpler networking options, compatible with Web
browsers and USB cables and are more suited to projects that require
taking inputs from external sources like sensors and actuators. Use
cases for this board includes light system and smart irrigation system.
The limitations of Beagle Board includes not suitable for complex
multimedia and Linux-​based projects due to unavailability of the graph-
ical and audio capabilities that the Raspberry Pi offers. BeagleBone
Black does not have a larger community support. BeagleBone Black
is a much costlier selection for add-​ons and peripherals. Table 2.2
presents the comparison of various development boards.

TABLE 2.2
Comparison of Development Boards
S. No Parameter Arduino Raspberry Pi BeagleBone
1 Example controller 8 bit 32 bit 32 bit
2 Flash code 32 kb SD,MMC card 4GB e-​MMC
3 RAM size 2 kb 256 Mb 512 Mb
4 Microcontroller Atmega3280P BCM2835 AM335x
chip
5 Clock 16 MHz 700 MHz 1 GHz
6 Operating system None Linux Linux,
Android
7 Price 20$ $35 $50
38 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t

REFERENCES
1. Snellma Henrik, Savolainen Mikko, Knaappila Jere, and Rahikkala
Pasi, Bluetooth ® 5,Refined for the IoT. Available online: www.
silabs.com/​documents/​public/​white-​papers/​bluetooth-​5-​refined-​
for-​the-​IoT.pdf, 2019.
2. Zigbee Alliance, Zigbee Specification. Available online: https://​
people.ece.cornell.edu/ ​ l and/ ​ c ourses/ ​ e ce4760/ ​ F inalProjects/​
s2011/​kjb79_​ajm232/​pmeter/​ZigBee Specification.pdf, 2008.
3. EnGenius Technologies, Wi-​ Fi Beacon Frames Simplified.
Available Online: www.engeniustech.com/​wi-​fi-​beacon-​frames-​
simplified/​, 2017.
4. RFC 8025: IPv6 over Low-​Power Wireless Personal Area Network
(6LoWPAN). Available online: https://​tools.ietf.org/​html/​rfc8025.
5. R. Zeqiri, F. Idrizi, and H. Halimi, 2019. “Comparison of
Algorithms and Technologies 2G, 3G, 4G and 5G,” 3rd
International Symposium on Multidisciplinary Studies and
Innovative Technologies (ISMSIT), Ankara, Turkey, pp. 1–​4.
6. LoRaWAN Alliance. Available online: https://​lora-​alliance.org/​
about-​lorawan.
7. A. Lavric, A. I. Petrariu, and V. Popa. “Long Range SigFox
Communication Protocol Scalability Analysis Under Large-​Scale,
High-​Density Conditions,” IEEE Access, 7, 35816–​35825, 2019.
8. R. Jain, “Constrained Application Protocol for Internet of Things,”
International Journal of Engineering and Technology, 857, 1–​
12, 2014.
9. C. Cho, J. Kim, Y. Joo, and J. Shin, “An Approach for CoAP-​
based Notification Service in IoT Environment.” International
Conference on Information and Communication Technology
Convergence (ICTC), Jeju Island, Korea, 2016, pp. 440–​445.
10. D. Ugrenovic and G. Gardasevic, “CoAP protocol for Web-​
based monitoring in IoT healthcare applications,” 23rd
Telecommunications Forum Telfor (TELFOR), Belgrade, Serbia,
2015, pp. 79–​82.
11. “MQTT.” Available online: http://​mqtt.org/​.
12. P. Dhar and P. Gupta, “Intelligent Parking Cloud Services Based
on IoT Using MQTT Protocol,” in International Conference
on Automatic Control and Dynamic Optimization Techniques
(ICACDOT), Pune, India, 2016, pp. 30–​34.
13. D.-​H. Kang et al., “Room Temperature Control and Fire Alarm/​
Suppression IoT Service Using MQTT on AWS,” in International
Conference on Platform Technology and Service (PlatCon), 2017,
pp. 1–​5.
 Internet of T h i n g s A pp l ica t i o n D e v e l o p m e n t 39

14. C. Trieloff et al., “AMQP Advanced Message Queuing Protocol

Specification License.”
15. “XMPP | An Overview of XMPP.” Available online: https://​xmpp.
org/​about/​technology-​overview.
16. J. E. Luzuriaga, M. Perez, P. Boronat, J. C. Cano, C. Calafate, and
P. Manzoni, “A Comparative Evaluation of AMQP and MQTT
Protocols Over Unstable and Mobile Networks,” in 12th Annual
IEEE Consumer Communications and Networking Conference
(CCNC), Las Vegas, NV, 2015, pp. 931–​936.
17. D.G. Costa and C. Duran-​ Faundez, Open-​ Source Electronics
Platforms as Enabling Technologies for Smart Cities: Recent
Developments and Perspectives. Electronics 7, 404, 2018.
 3

IOT–​C LO UD CO NVER GEN CE

3.1 INTRODUCTION
Chapters 1 and 2 have given us a vivid picture about IoT and the
associated design issues and challenges. The chapters have also given
us an insight into the IoT application development process, key tech-
nologies used, and the platforms available that support in the actual
development of the applications. They have built a strong founda-
tion for the rest of chapters in this book in which we will discuss
about the importance of Cloud in IoT and the various approaches and
techniques that can be applied in real-​time scenarios encountered
when developing an IoT application.
In this chapter, we primarily discuss how Cloud plays a vital
role in realizing IoT to its full potential. The chapter is organized as
follows. Section 3.1 discusses the opportunities and challenges that
are encountered when the IoT–​Cloud convergence comes into pic-
ture. Section 3.2 provides a bird’s eye view on the architecture for
convergence. Section 3.3 elaborates on data offloading and computing
from the IoT perspective. Section 3.4 explains the concept of dynamic
resource provisioning for IoT. Section 3.5 introduces the various test
beds and technologies with respect to IoT–​Cloud Convergence.
The topic under discussion being IoT–​Cloud Convergence cannot
be well explained without providing an insight into the basics of the
Cloud technology in brief. The National Institute of Standards and
Technology (NIST) defines Cloud computing as “a model for enab-
ling ubiquitous, convenient, on-​demand network access to a shared
pool of configurable computing resources (e.g., networks, servers,
storage, applications, and services) that can be rapidly provisioned
and released with minimal management effort or service provider
interaction” [1]. In simple terms, Cloud computing is on-​demand
delivery of scalable computing resources (storage, servers, etc.)
provided as a service to the users over the internet, thereby allowing

41
42 I o T- C l o u d C o n v e r g e n c e

flexibility and efficient scaling according to one’s business needs.

As the Cloud computing model works on the ‘pay-​as-​you-​go’ model
(pay for only as much as you use), it helps greatly in reducing oper-
ational and infrastructure maintenance costs. Also, as the infrastruc-
ture setup and maintenance are taken care of by the Cloud service
providers (Amazon, Google, Azure etc.) businesses can now concen-
trate on developing innovative products and applications pertaining to
their lines of work.
The above definition most precisely explains Cloud com-
puting and its usage. However, a quick glance into the quintessen-
tial characteristics of Cloud computing, Cloud computing service
models, and Cloud computing deployment models will provide the
readers with a better understanding on why Cloud computing plays
and integral role in IoT.

3 . 1 . 1 Fi ve Q uin tesse n tial Ch arac ter is t ics o f

Clo u d Co m putin g

On-​demand self-​service: This refers to the facility of provisioning

resources (servers, virtual machines, and databases) as and
when the user requirement arises without the necessity of any
human interaction or the service provider.
Broad network access: This feature refers to the uniform avail-
ability of resources over the internet to a broad variety of
devices via standard access mechanisms.
Multi-​tenancy and resource pooling: Multi-​tenancy refers to the
availability of sharing the Cloud computing resources with
multiple users and customers without any disruption/​intrusion
to their privacy and security. Resource pooling is a feature that
enables multi-​tenancy to be achieved seamlessly, by dynam-
ically reassigning the various physical and virtual resources
among multiple users.
Rapid elasticity and scalability: Cloud computing provides
the users with the facility of scaling up or scaling down the
resource capacities depending on the dynamically changing
business needs.
Measured service: The utilization of resources can be monitored
and tracked, which provides a level of transparency to the users.
Users and the service providers can keep an account of the
activities performed with the resources and billed accordingly.
 I o T- C l o u d C o n v e r g e n c e 43

3 . 1 . 2 Clo u d Computin g Se r vice Mo dels

The Cloud computing service models are of three types.

Software as a Service (SaaS): The SaaS model allows users

to use any Web-​based application without any requirement
to download or install the application/​software on the local
infrastructure. The entire computing stack is managed by the
service provider. Examples include Dropbox and Microsoft
Office 365.
Platform as a Service (PaaS): The PaaS model enables the
customers to develop and test various applications for their
business needs by providing them with a virtual runtime envir-
onment. This eliminates the need to worry about setting up the
appropriate operating system (OS), middleware, and runtime
environment on every individual machine. Along with the vir-
tual runtime environment, other resources such as the storage,
servers, and networking are taken care of by the service pro-
vider. Examples include Google App Engine and Amazon Web
Services (AWS) Elastic BeanStalk.
Infrastructure as a Service (IaaS): The IaaS model provides the
users with the entire infrastructure such as the servers, storage,
and networking that is required for business operations. This
eliminates the necessity for customers to worry about the infra-
structure procurement, setup, and maintenance costs.

3 . 1 . 3 Clo u d Computin g De ploymen t Mo dels

The Cloud computing deployment models are of four types:

Private Cloud: The private Cloud is an infrastructure that is

hosted for use by a single organization or company. This type
of hosting provides a great level of security; however, the cost
of maintaining is higher.
Public Cloud: The public Cloud is an infrastructure that is hosted
by a service provider and made available to customers publicly
over the internet based on a shared cost model.
Community Cloud: The community model is an infrastructure that
is shared among companies/​organizations that fall under a par-
ticular domain such as banks and government organizations.
44 I o T- C l o u d C o n v e r g e n c e

The infrastructure can be managed internally by the community

or by a third party.
Hybrid Cloud: Hybrid Cloud is an infrastructure that is a com-
bination of both the private and public models. This model is
brings the best of the two models in terms of security, scal-
ability, flexibility, and cost effectiveness.

After a basic discussion of the Cloud technology, we now move into

the actual topic of discussion: IoT–​Cloud Convergence.

3.2 OPPORTUNITIES AND CHALLENGES

3 . 2 . 1 Io T Requireme n ts to M e et th e Fut ur e
Ma rket P o tential
The Internet of Things (IoT) is basically an orchestration of devices/​
machines that are provided with the capability of uniquely identi-
fying and communicating with each other in order to carry out a task
without the need for any human interference. The IoT in its course of
development has started evolving from small standalone applications
toward a vision of achieving a completely connected world. To
achieve this futuristic vision, many critical aspects and requirements
must be taken into account.

• With numerous heterogeneous and mobile devices being con-

stantly connected over the internet, ubiquitous access and
uninterrupted connectivity should be ensured for providing
uninterrupted service.
• An efficient means to dynamically collect, store, and process
the huge volumes of data that are collected or produced by
numerous devices and users is required.
• The Internet of Things also has end users interacting with the
edge nodes/​sensors, which collect user-​sensitive information
pertaining to the application’s usage as in the care of a health
care system. Such sensitive data needs to be guaranteed privacy,
security, and reliability.
• Also, each real-​time application has different types of users who
will look for specific services in the application. Supporting
user preferences based on the context is also vital to the success
of the application.
• An application can have a wide deviation in the number of
users/​service requests that it handles at different points of time.
 I o T- C l o u d C o n v e r g e n c e 45

Scalability of resources, enabling them to grow or shrink based

on business needs, is a very import feature that ensures cost
effectiveness.

A simple scenario can help the reader gain good understanding of the
above-​mentioned requirements. Consider a multi-​specialty hospital
handling thousands of inpatients and outpatients in a day. A person
looking for consultation/​ treatment books a time slot to visit the
hospital. The application basically requires consent from the user
enabling the hospital authorities to access his medical records and
photo ID. The app also records the user’s registered phone number.
This collection of user-​sensitive information must be done without
affecting the privacy of the user. Also, the collected information,
being highly sensitive user medical records, need to be stored with
utmost security measures in place.
When the user enters the premises, the CCTV at the entrance
recognizes him and gives him real-​time guidance and notifications
on his registered mobile device. Following the instructions, the user
is dynamically assigned to an available doctor for consultation. Any
further tests that are recommended are uploaded in the app and the
user is dynamically directed to the appropriate labs for the tests. The
results are made available to the consulting doctor through the app.
The doctor then analyzes the reports and specifies the medication in
the app itself. The user then goes to the pharmacy and authenticates
him with the registered phone number. Upon successful authen-
tication, the medical prescription alone is made available to the
pharmacist who provides him with required medicines and leaves
the hospital.
The entire scenario that has been narrated from the context of
a single user it has to be done for thousands of users per day. The
data collected from all such users can be huge and needs a separate
dedicated secure space for storage. Also, the processing of lab results
and assignment of doctors to users based upon the availability must
be done dynamically, which will amount to significant computational
overhead. Also, the number of users who enroll for a visit will vary
hugely with respect to time. During the day, we might have many
users visiting the hospital and the number decreases at night. There is
no point in keeping all the instances (storage, processor, and virtual
machines) up and running when the demand is low. To minimize the
cost and use the resources effectively, the resources should be scal-
able according to the needs.
46 I o T- C l o u d C o n v e r g e n c e

3 . 2 . 2 Ho w Cloud Come s in Han dy ?

The scenario is Section 3.1.1 has clearly explained the requirements
of IoT to meet the future market potential. Earlier in the chapter, we
also discussed the five quintessential characteristics of Cloud, its ser-
vice models, and deployment models. With these in mind, it is quite
clear that Cloud is the appropriate candidate that will meet all the
requirements and challenges that the IoT might face in its journey
toward establishing a connected future world.
In small standalone consumer applications like the smart home and
care for disabled and elders, the requirement for capabilities such as
processing power, storage space, and computing can be facilitated by the
participating devices (mobile phones, Raspberry Pi kits) themselves.
The resource requirements for such applications are low. However,
for larger commercial real-​time applications such as health care and
transportation, the demand for resources is huge due to the enormous
data that is collected and also the complexity of computations that are
involved in the system. The demand for resources (storage, servers
and processors) in such huge applications many a times cannot be sat-
isfied by the participating devices alone. This is where Cloud comes
in handy by catering to larger business needs.
The fundamental requirement of IoT is resource sharing. Cloud
computing not only works on the principle of sharing resources but
it also emphasizes on maximizing the availability of resources. One
of the key features of Cloud is the virtualization of underlying phys-
ical devices, which makes it easier for users to share the same hard-
ware in a seamless manner. IoT has many mobile users and devices.
Services should be made available to these devices/​users independent
of the location. Cloud being location independent allows services to
be accessed from anywhere and from any device over the internet.
The above discussion assures that the convergence of IoT and Cloud
can lead to many successful developments in the field.

3 . 2 . 3 Ap p li cation A reas f or IoT –​Clo ud

Co nverg ence
In an interconnected world with billions of IoT devices operating
worldwide, data is being captured in abundance. To derive meaning of
this data, it is usually processed in the Cloud and the outcome is used
to improve process efficiencies, explore new revenue streams, and
generate new insights into the existing business strategies. Whether it
 I o T- C l o u d C o n v e r g e n c e 47

is sports analytics, health and safety at hospitals, reducing congestion

and improving safety at airports, customer satisfaction in shopping
malls, reducing traffic congestion in cities, or detecting component
defects at manufacturing facility to save human lives, or even improve
and fix artifacts in the photo captured in your mobile phone in real
time, without the confluence of IoT and Cloud, nothing would be pos-
sible. It is predicted that early adaptors of Edge-​enabled technologies
will benefit the most.

3.2.3.1 IoT and Cloud Confluence in Agriculture

IoT devices are revolutionizing agriculture in the ways we have not
imagined. Autonomous robots in a swarm multitask agricultural activ-
ities like weeding, seeding, harvesting, and hauling. These bots are
modular and always in use. In addition to running on electricity, they
are lighter in weight and cost less to operate than traditional agricul-
tural equipment. These bots have multiple cameras and lidar sensors that
run in parallel, which enables them to navigate in any direction. They
run neural networks downloaded from Cloud and can identity rows of
crops and follow them accurately. In fact, they can accurately identify
weed from crop and remove them. These bots communicate with each
other via wireless mesh networking and the learning can be transferred
between each other. In near future, farms will be run autonomously
rather than renting it out. This in turn is a huge saving for the landowner.

3.2.3.2 IoT and Cloud Confluence in Manufacturing

In today’s age of manufacturing, the processes are extremely complex,
precisely engineered and infinitely faster than what it was a decade
ago. Every activity is timed to a second. This is mostly accomplished
by automation using robotic arms and rule-​based anomaly detection
systems. Rule-​based anomaly systems are legacy systems that lack
the modern Cloud computing power and IoT.
Image-​based anomaly detection systems in a high-​yield manufac-
turing environment can process tens of millions of images per day and
perform inference at multiple points per second, which cuts manufac-
turing time by days or even weeks on a longer run. This also helps
detect more defects before entering the final assembly line. Image-​
based anomaly systems usually have high cameras over the edge
devices where inference happens, and the training happens on an on-​
premises Cloud. Manufacturing industries prefer on-​premises Cloud
due to sheer volume of data needed to be sent online to the Cloud for
processing.
48 I o T- C l o u d C o n v e r g e n c e

3.2.3.3 IoT and Cloud Confluence in Oil and Natural

Gas Rig Safety
IoT and Cloud play a vital role in ensuring safety of people working
in dangerous environments. A combination of mechanical and human
errors has led to loss of lives in the past and present. AI-​powered
computer algorithms powered by on-​premises Cloud and network of
cameras sensors to locate workers at unsafe locations and alert them
before any accidents could happen.

3.2.3.4 IoT and Cloud Confluence in the Retail Industry

IoT and Cloud play a vital role in the retail industry. They can be
gainfully employed to cut operational expenses, improve decision
making, and increase revenue by billions of dollars. On an average,
retailers lose 2 percent of sales due to shrinkage. For large retailers,
this shrinkage would cost billions of dollars in losses. IoT devices
backed up by AI in the Cloud can accurately detect mishaps like mis-​
scans using intelligent video analytics. These systems are proven to
be 99 percent accurate and immediately inform the sales associates so
that they can take appropriate action. IoT-​powered store analytics tells
retailers about popular store locations, unique visitors, and popular
hangout areas in the store. It also lets the retailers know about the cold
spots. Empowered with this information, retailers can do precision
marketing and do better store merchandizing.
Autonomous shopping, also known as grab-​ and-​
go stores, is
gaining popularity. This kind of shopping enables customers to
experience faster check-​outs without a cashier by just using their
mobile phones. This kind of stores reduces operational expenses and
delivers a faster shopping experience.

3.2.3.5 IoT and Cloud Confluence in Supply Chain and Logistics

Supply chain and logistics is one of the most complex processes. It
involves accurate and fast forecasting. Companies save millions of
dollars if the supply chain gets shortened even by one day. Warehouse
logistics involves cutting-​edge robots with computer vision over the
edge and the power of Cloud to compute millions of items by store
combinations, thereby improving retail forecasting.

3 . 2 . 4 Cha lle n ges T h at C ome with C o n v er gen ce

The previous sections have justified that IoT–​Cloud convergence is
undoubtedly the best fit for the implementation of many real-​time
 I o T- C l o u d C o n v e r g e n c e 49

IoT applications. However, there are a few challenges that might be

encountered while combining the two technologies. Addressing these
technologies and developing countermeasures will help in the devel-
opment of effective real-​time applications.
Secure communications and transfer of data: A secure data
transfer between the service providers and devices/​things is neces-
sary to ensure secrecy and integrity of data. A careful choice of crypt-
ography algorithms and appropriate protocols plays a very critical
role in protecting the data in transit from being interpreted and/​or
tampered by eavesdroppers.
Proper identification and authentication of devices/​ things and
Cloud service providers: Given the fact that an IoT system/​application
can have encounters with devices that it has never experienced before,
careful mechanisms should be devised to properly identify, authenti-
cate, and authorize the devices to ensure security against intruders.
Also, the Cloud service providers being third party providers need to
establish their identities and authenticate themselves to the devices/​
things in the instances where sensitive data is being shared for com-
putational and storage purposes. Therefore, a two-​way authentication
is required in order to ensure a fool proof system.
Privacy of user’s information/​data needs to be ensured by using
proper access control mechanisms, which determine the level of
access that each participating entity in the system has to the data.
Data is made available based on the “need to know” principle and the
type of work that needs to be performed by the entity. Not all data is
always required for all computations. The Cloud in most situations is
also accountable for regulating and coordinating the devices/​things
in the system with proper access control policies and mechanisms.
Also, there might be certain situations where a particular entity
is unavailable to perform the assigned task and the task has to be
delegated to another entity dynamically. For example, in a situation
where a surgeon is unable to perform a surgery due to an unforeseen
situation, there must be a mechanism to auto delegate the task to a
surgeon with a good track record. This computation should be done
dynamically, and assignment should be done with no or little lapse
of time.
Choice of appropriate Cloud deployment models is very critical
for the success of the business. For example, an application that deals
with highly sensitive data cannot rely on the public Cloud platform
to store the data; however, if the application uses some high-​end
resources for certain computations (e.g., Graphics Processing Unit
50 I o T- C l o u d C o n v e r g e n c e

[GPU]) that are too costly, then public Cloud comes in handy. In such
a situation the organization must find a balance between the private
and public Cloud (hybrid Cloud) such that it achieves cost effective-
ness without any compromise in terms of security and performance.
The Cloud service provider, being the aggregator of information/​
data, should be accountable for the data and abide by the privacy
terms and conditions. Organizations must make a careful choice of
the service providers based on the service providers’ reliability and
efficiency in providing services.
Virtualization of resources: Advanced and specialized hardware
resources such as GPUs should be effectively virtualized such that
many virtual machines share one physical GPU (cost-​effective). The
inverse of this situation is also possible wherein multiple virtual
GPUs are allocated to a single virtual machine that executes highly
demanding workloads. Also, with the hugely changing demands in
an IoT application, the need for resources can increase suddenly and
fall sharply. Under such situations dynamic allocation of resources
must be done in order to ensure uninterrupted services. Virtualization
enables the Cloud to maintain a pool of resources and allocate the
resources as per the changing demand.
A seamless transition of resources is crucial considering the ubi-
quitous nature of participating devices. When a device moves from
one location to another, resources might have to be reallocated in
order to felicitate better user experience in terms by reducing the
latency involved. Also, if a particular resource (e.g., server) in a geo-
graphical location goes down, the services should be continued with
fall back resources.
Considering the huge number of devices that will be connected
over the internet to the Cloud (devices can also span across geo-
graphical locations), and the huge volumes of data that they gen-
erate from various sources, the Cloud should be equipped to handle
the demand that might arise in terms of storage and computational
power. The Cloud should be capable of decentralized monitoring and
maintenance of resources. The Cloud should also be able to handle
and organize the data coming in from multiple devices mostly in an
unstructured or semistructured format. Whenever any computation or
even simple storage of the data is required, it should be converted
into a structured format to enable further analysis via application of
algorithms and logic.
Resource-​intensive computations when encountered in an IoT
application can leverage the Cloud resources to complete the task
 I o T- C l o u d C o n v e r g e n c e 51

and a faster and cost-​effective manner. All that needs to be one is to

transfer the data (data offloading) to the appropriate Cloud resources
for computation; however in real-​time applications, the complexity
is increased due to heterogeneity of platforms and devices, different
formats of data collected from different sources, unpredictable net-
work conditions, and unique characteristics of the application.
The above are a few typical challenges that can be encountered in
real-​time when the IoT–​Cloud convergence comes into the picture.
Addressing these issues by incorporating effective mechanisms can
help develop strong and resilient applications that can lead to the real-
ization of a future connected world.

3.3 ARCHITECTURE FOR CONVERGENCE

3 . 3 . 1 An O ve r view of Cloud- ​b ased I o T S er v ices
There are numerous Cloud service providers in the market who offer
services for businesses to develop productive, scalable, and real-​time
IoT solutions. Most of the service providers offer specified services
that help to meet the requirements and growing needs of IoT as
discussed in Section 3.1.1. This section will provide a quick over-
view of the IoT platforms/​services provided by major Cloud service
providers such as Amazon, Google, and Microsoft.

3.3.1.1 Amazon Web Services (AWS) IoT Services

AWS offers numerous services that help customers in developing
reliable and resilient IoT solutions in various domains. Security,
scalability, data management, and analytics are the primary features
offered by its services. The AWS IoT services are broadly divided
into three major categories: device software, connectivity and control
services, and analytics services.

Device Software
The device software enables provides the users with hassle-​free experi-
ence in connecting the devices/​things to the Cloud and equips the
devices (also called as edge nodes) to be able to perform a number of
critical computations. The two services offered under this category are:

• FreeRTOS is an operating system that is provided by AWS that

helps in easy programming, deployment, and management of
the small-​sized low-​end edge devices. It also eases the process
of securely connecting the edge devices to the Cloud.
52 I o T- C l o u d C o n v e r g e n c e

• AWS IoT Greengrass is a software that enhances the capabilities

of edge devices to minimize the dependency on Cloud for all
computations. It also ensures minimized latency and overhead
in transporting the data over to the Cloud and vice versa by the
data caching powered at the edges.

Connectivity and Control Services

The services offered in this category ensure that devices abide by
specified security protocols, eases the process of registry, and con-
nectivity of devices/​things over the internet to the Cloud. The ser-
vices also make it possible to manage and orchestrate the multitude
of devices through the Cloud. Following are the services offered in
this category.

• AWS IoT Core helps in easing the communication between

devices and Cloud by ensuring security in the process.
• AWS IoT Device Defender ensures that the connected devices
abide by the prescribed security guidelines by constantly
monitoring the devices.
• AWS IoT Device Management enables proper registration,
orchestration, and management of devices with the Cloud
performing the job of a mediator.

3.3.1.2 Analytics Services

The growing number of heterogeneous IoT devices contributes to
huge volumes of unstructured data being generated, which poses a
greater challenge for the applications to process and analyze such
data. AWS provides services that makes tedious jobs much easier for
customers. Services provided under this category are as follows.

• AWS IoT Analytics is a service that enables the users to easily

run complicated analytics on huge volumes of data generated/​
collected by the devices.
• AWS IoT SiteWise service eases the process of gathering, organ-
izing, and analyzing data that comes in at a larger scale.
• AWS IoT Events helps in easing the process of event detection
and response that originates from the large number of sensors
and applications.
• AWS IoT Things Graph eases the connection establishment
activity between the IoT devices and the Cloud services.
 I o T- C l o u d C o n v e r g e n c e 53

3.3.1.3 Microsoft Azure IoT Products and Services

The Microsoft Azure IoT products and services enable the connection
and management of numerous IoT devices. It offers a scalable and
open platform for the development of Cloud-​based IoT applications.
The services offered by Azure are as follows.

• Azure IoT central enables faster connection of IoT devices

to Cloud, offers centralized management of the devices, and
provides application programming interfaces (APIs) and
connectors that establishes direct connection between the data
generated/​collected by IoT devices and the decision-​making
services.
• Azure IoT solution accelerators offer predefined templates or
application code for common scenarios like remote monitoring
and device simulation, thereby enabling customers build
customized applications using the off-​the-​shelf code and scale
them according to their business requirements. This helps in
speeding up the development process and saves effort that can
be used for innovation.
• Azure IoT Edge helps to move certain Artificial Intelligence
and analytics workloads to the edge devices, thereby pro-
viding near real-​ time responses to requests, reduced costs,
minimized latency, and allowing offline operations or intermit-
tent connectivity.
• Azure IoT Hub offers support for secure and authenticated two-​
way communication between the IoT devices and the Cloud
(Azure).
• Azure Digital Twins helps to develop a visual representation or
model of the physical environment with proper representations
of communication between people, places, and devices. It
facilitates data collection from the physical environment via
models rather than relying on sensors.
• Azure Time Series Insights enables businesses to convert the
enormous data produced over time from sensors into mean-
ingful data. It stores time series data in near real time and
provides insights from this data in minutes without any pre-​
preparation or efforts spent in coding. Consumers gain a global
view of what is happening with their sensors and machines.
• Azure Sphere helps in protecting the data, infrastructure and
helps maintain privacy. It provides security for the already in
54 I o T- C l o u d C o n v e r g e n c e

place devices by means of updates and provides consumers

with new IoT devices that are highly secure. It also provides
certified chips with built in security measures. The Azure
sphere OS provides a trustworthy platform for IoT application
development.
• Azure Maps helps incorporate location intelligence into IoT
applications. It provides facilities like search, routing, traffic
insights, and provides real-​time (mobility) location intelligence.
• Azure RTOS is a real-​time operating system that eases the con-
nectivity of embedded devices and gateways to the Cloud. It
helps in developing easy-​to-​use applications and enhances the
real-​time performance of simple, low-​capability devices.

3.3.1.4 Google Cloud-​based IoT Services

Google Cloud IoT offers multiple tools and services that aid devices
to connect to the Cloud. It likewise offers services and tools to gather,
store, and process the data that is collected. Google Cloud IoT plat-
form is completely scalable, having fully managed Cloud services.
Its software stack is designed to address both edge/​on-​premises com-
puting with AI infused at all levels.

Google Cloud IoT Core is a fully managed service that allows

easy and secure connection and management of IoT devices. It
also allows ingesting data from millions of IoT devices across
the globe. Clout IoT core with other services on the Cloud pro-
vide complete solution, end-​to-​end retrieval, and visualization
of IoT data in real time.
Google Cloud Dataflow is a service for streaming and batch ana-
lysis. It enables fast and simplified streaming data pipeline with
lower latency. Dataflow automates provisioning of resources
and manages them, thereby offering infinite capacity to auto-
matically manage seasonal or spiky workloads.
Google Cloud ML Engine is used for training, deploying, and
running machine learning (ML) Models. Cloud ML Engine
makes it less complex for machine learning developers and
data scientists to take their ML projects from deployment to
production. The AI platform supports Kubeflow, which lets
portable ML pipelines that can be run on-​ premises or on
Google Cloud.
Google Cloud Pub/​Sub is a service that, for Ingest connection
and management, provides reliable staging location for data on
 I o T- C l o u d C o n v e r g e n c e 55

its journey to processing. The Pub/​Sub takes care of scaling,

load isolation, and expansion of applications and pipelines to
new regions. It also simplifies the development of event-​driven
microservices, allowing services to react to it.
Google BigQuery is a serverless, scalable Cloud data warehouse
for fast querying. It can analyze petabytes of data using ANSI
SQL at high speeds and no operational overheads.

3 . 3 . 2 S ta te of th e A r t: Con ve rge n ce
Archi tectures
In recent times, there has been an ever-​increasing demand for IoT
applications. However, the real-​time challenges that are being faced
in realizing the IoT to its full potential have led way to a consider-
able research in the area. With advancements in Cloud computing,
researchers have started to see the great benefits that can be obtained
by bringing IoT and Cloud together. Much research has been done in
developing an apt architecture that would put Cloud to a better use in
IoT applications.
In reference [2], the authors have proposed a simple architecture to
depict the future Internet of Things. The Cloud has been incorporated
into the system to function as a storage platform to store all the infor-
mation collected/​generated by the devices. It also functions as a medi-
ator/​managing authority that protects stored data by implementing
access control rules and policies. Also, it serves as a computa-
tional platform wherein all the huge power-​consuming calculations
related to access management are done. Also, the paper discusses a
decentralized mechanism to help in the registry and monitoring and
management of devices/​things that are connected to the Cloud.
In reference [3], the authors have proposed a generalized IoT–​
Cloud architecture to meet the demands of IoT applications. With
millions of devices being connected over the internet and demanding
application-​specific services, it becomes difficult to manage and serve
all such requests. The IoT Cloud proposed in this paper emphasizes
the virtualization of resources in order to ensure full utilization of
physical machines (improved resource utilization) and scalability
of servers. A virtual resource pool (virtualization technique using
hypervisors) is established on numerous physical machines with all
the required hardware resources. Based on the demand, the resources
from the pool can be assigned to the virtual machines dynamically.
56 I o T- C l o u d C o n v e r g e n c e

The system also includes other supporting components such as

databases, application server, and load balancers to ensure availability
and improved performance throughput. Additionally, HTTP and
MQTT servers discussed in Chapter 2 have been used to provide ser-
vices to end users and to ensure real-​time communication among the
connected devices, respectively.
In reference [4], the authors have presented a Cloud-​based IoT
architecture, which basically consists of three layers, namely, the
application layer, perception layer, and the network layer. The per-
ception layer consists of protocols and mechanisms that identifies the
various objects and collects information from the environment. The
network layer consists of protocols that help in smooth transmission
of the collected/​generated data over to the internet/​Cloud. The appli-
cation layer is the one which regulates the access to various services
based on demand.
In reference [5], the authors have proposed an IoT–​Cloud architec-
ture that presents a model for device/​things monitoring and control.
The architecture is basically composed of three layers: the wireless
device network, gateway, and the Cloud service. The 6LoWPAN
technology discussed in the previous chapter has been used to imple-
ment wireless device network, which collects data from the various
connected devices. The gateway transforms the collected data to fit
the protocol of the Cloud service. The Cloud performs the job of the
mediator with the help of the event manager and event rule manager
services that help authoring of rules and regulating access to services.
In reference [6], the authors discuss a multi-​Cloud IoT architec-
ture called the iKaaS architecture. The architecture introduces the
concept of local Cloud and global Cloud to provide an intelligent,
privacy-​preserving and secure big data analytics platform. The local
Cloud has enough resources and capabilities (computing, storage, and
network) to service requests from users/​devices within a specified
geographical region over a specified time period. The global Cloud
serves the conventional purpose of providing dynamic and demand
computation and processing capabilities. It helps to provide increased
business opportunities in terms of scalability, reliability, efficiency,
and, most importantly, ubiquity.
In reference [7], the authors have proposed a data-​centric IoT
framework. The proposed framework consists of two parts: local field
and Cloud platform. The local field encompasses the personal area
network, local area network, and the wide area network implemented
 I o T- C l o u d C o n v e r g e n c e 57

using the Thread, WiFi, and the Long Range (LoRa) communication
protocols that help in establishing the connection among the partici-
pating devices and gateways. The Cloud platform is implemented
using the Azure Cloud platform services (discussed in Section 3.2.1)
enabling provisioning and management of devices/​things, storage,
analytics, and visualization.
In reference [8], the authors have presented a novel IoT–​Cloud
architecture for “Big Stream” applications that have many data
sources, higher rates of information exchange, and greater need for
real-​time processing. It aims at minimizing latency and improving the
resource allocation process with the help of a listener-​oriented com-
munication model. The services or applications that are interested in
specific data from a data source (edge devices) register themselves
and wait upon them for any input. Based on the application’s specific
needs mentioned as rules, certain specific Cloud services work upon
the data (normalize the data) in converting it into a format applicable/​
desirable to the consumer application/​service.
In reference [9] the authors have proposed an edge-​based IoT
Cloud architecture with the aim of reducing latency and improving
the security and efficiency of the system. To enhance the security
aspect in the Cloud and enable IoT, the authors have adopted the trad-
itional trust-​based evaluation mechanism. A large part of the heavy
resource-​consuming trust calculations is done at the edge networks.
This helps in reducing the latency and cost incurred in transferring the
information back and forth between the Cloud and devices.
In reference [10], the authors have presented an architecture
called the Internet of Things edge Cloud federation. The architec-
ture addresses the problem faced in the transfer of data between the
devices and Cloud due to the different protocol stacks and hetero-
geneity of devices. The model also uses the apache kafka publish/​
subscribe platform for fault tolerance and Kubernetes that provides a
pipeline for data processing.
In reference [11], the authors have presented a first of its kind
IoT architecture using osmotic computing. This computing tech-
nique involves a mechanism to manage the various resources, data,
and services as the elements move across various infrastructures that
are heterogeneous in nature. In order to orchestrate or manage such
resource movements, the system has to be trained using an Artificial
Intelligence module.
58 I o T- C l o u d C o n v e r g e n c e

3 . 3 . 3 A S i mplif ie d C on vergen c e M o del

Figure 3.1 depicts a simplified IoT–​Cloud convergence model. The
architecture is built upon the most popular three-​tier architecture for
IoT, which includes the IoT devices layer making up the bottom or
first layer. The second layer comprises one of the edge-​computing
technologies (mobile edge computing, fog computing, Cloudlets,
etc.), which will be discussed in Section 3.3.3. The last layer consists
of the Cloud layer. The edge computing layer helps to bring some
of the services offered by the Cloud closer to the IoT devices layer,
thereby helping to minimize cost and latency, and improve privacy,
security, and efficiency of the application.
The data normalization layer lies above the edge computing middle
layer, which houses several protocol connectors that act as a pipe-
line and helps in organizing the heterogeneous data coming in from
multiple platforms into a format that is understandable to the Cloud
resources, enabling it to process the data at ease and give accurate
results. It acts as a bridge/​mediator between the IoT device layer and
the Cloud.
The Cloud apart from the various resources that it houses also has
a set of supporting services that assist in effective resource allocation.

Figure 3.1 A simplified IoT–​Cloud convergence architecture.

 I o T- C l o u d C o n v e r g e n c e 59

The device registry is responsible for maintaining a record or old and

new devices that the Cloud encounters requests from. This enables to
establish a mutual authentication between the Cloud and the device,
thereby reducing the possibility for any security threats and attacks.
The Cloud also has a resource pool, which houses several virtual
machines that are replicas of the underlying physical machines. The
resource pool has a scheduler that is responsible for dynamically allo-
cating or de-​allocating a virtual resource to a specific task.

3.4 DATA OFFLOADING AND COMPUTATION

3 . 4 . 1 Da ta Of f loadin g an d Compu t at io n : A n I o T
Persp ecti ve
The recent development in the field of IoT has led to a widespread
adaptation of its applications. The IoT has several low-​cost wireless
sensors and devices that are connected to each other to share and pro-
cess the collected data. The sensors and devices constantly monitor
their surroundings and collect huge volumes of data. With the incorp-
oration of Cloud to bridge the gap/​shortcomings of IoT in terms
of resources, Cloud computing has become an integral part of IoT.
However, this incorporation of Cloud to IoT to offer different services
means that there needs to be a huge volume of data transfer between
the connected IoT devices and Cloud. Although offloading of data
from devices to Cloud helps overcome the limitations of IoT, high
additional cost is incurred and maintenance is required for commu-
nication links.
The Cloud in most cases being located far away from the users
requires large amounts of data to be transferred to and from the
Cloud. This movement of data causes several performance issues like
heavy loading of servers, longer response times, lowered bandwidth
for communications, and increased costs. All these limitations have
led to an increasing demand in finding alternative data offloading
techniques. One of the common techniques includes setting up a ter‑
minal to terminal (T2T) network that totally depends upon the direct
communication between mobile nodes. This communication there-
fore does not require a backbone infrastructure for communication.
This is more suitable for applications that gather and use data that is
delay tolerant. Such time-​insensitive data can be withheld in a par-
ticular place before it can be transmitted over to a distant Cloud ser-
vice or infrastructure for analysis.
60 I o T- C l o u d C o n v e r g e n c e

Another alternative is the use of mobile data offloading technique

wherein the network has many mobile nodes. In situations where the
network is heavily loaded with traffic or is not accessible due to the
geographical location, the network provider may decide to send
the delay-​tolerant data to only a few mobile nodes within the speci-
fied range. Later, when these mobile nodes meet other mobile nodes
(that are interested in the content/​data), they exchange the data by
means of peer-​to-​peer communication (P2P). A variant of this tech-
nique is the prediction-​based offloading mechanism, which functions
based on the predictable behavior of human mobility. Studies have
shown that the mobility of human beings follows a particular pattern
and has a correlation to the locations forming clusters [12]. Based on
this theory, data is offloaded to mobile nodes that are more likely to
transfer the data to the required destination whenever the opportunity
arises (opportunistic peer-​to-​peer data exchange).
Although the above discussed mechanisms are suitable in most
cases, certain applications cannot afford the delay caused by adopting
them. Applications that collect and process time-​sensitive data conse-
quently work on strategies that improve the efficiency by offloading
the data, processing, and storage closer to the sources that collect
information/​data by adopting the edge computing technologies.

3 . 4 . 2 Ed g e Computin g Tec h n ologie s f o r an

Io T Netwo rk In f rastructure
Edge computing is a mechanism used to enhance the data processing,
computing, and storage activities in applications/​systems where the
Cloud is an integral part. The processing (computing and storage)
capabilities are brought closer to the networks’ edge (data sources),
thereby leading to several advantages.

• It helps in reducing the latency involved in transporting the data

to and from the Cloud.
• It also helps to perform computations offline without depend-
ency on the Cloud for all processing activities.
• The performance and throughput-​related challenges that occur
due to the intermittent internet connections in most cases can
also be overcome.
• Reduces the load on Cloud servers and helps reducing the load
on communication links (congestion) due to the transport of
huge volumes of data (bandwidth-​intensive big data).
 I o T- C l o u d C o n v e r g e n c e 61

• It helps in near real-​time data processing and analytics closer to

the data sources.
• It reduces the organizations’ operating costs and the scheduled
downtimes (from the Cloud providers).
Following are a few edge computing technologies. All the
below discussed technologies form the middle tier of the three
architecture, that is, the IoT devices, edge/​fog/​Cloudlets etc.,
and the Cloud.

Multi-​Access Edge Computing or Mobile Edge Computing

(MEC) is an architecture that is implemented to lower the
network congestion for improved performance of the appli-
cation. It also satisfies the applications’ needs such as higher
bandwidth, lower latency, and location awareness. The
architecture enables the resources that make up the Cloud
to be placed within the radio access network (RAN). The
MEC in most cases are developed using open source hard-
ware and software like SDN (software defined network) and
NFV (network function virtualization) that make use of the
facilities provided by Cloud and adopts the virtualization
mechanisms of the underlying hardware on which it runs.
The virtualization infrastructure is housed in the MEC’s
mobile edge host (MEH) layer. The MEH is the key compo-
nent of MEC that functions as the application server enab-
ling the movement of storage, computing, and processing
capabilities closer to the edge of the network.
Fog computing is an architecture that is implemented with
distributed data centers closer to the edge devices. Individual
service providers build distributed data centers or distributed
Cloud, which houses various Cloud computing resources.
The resources that offer services can be placed in a cen-
tral office or within the customer premises also. It is more
like MEC architecture in bringing the Cloud capabilities
closer to the network’s edge. However, in fog computing, a
common centralized computing server processes all the data
from various endpoints in the network, while in edge com-
puting, every network performs data processing. Another
major difference lies in the positioning of the intelligence
and computation capabilities. Edge computing moves all
or most of the processing and computational capabilities
into the edge devices, while in fog computing the pro-
cessing capabilities and intelligence are pushed into a local
62 I o T- C l o u d C o n v e r g e n c e

area network (LAN) that houses various fog nodes. The

fog nodes receive the near real-​time data and perform ana-
lysis on them (for time-​sensitive data) or forward the data
to the Cloud for further action (for time-​insensitive data)
depending on the use case and nature of data received. Fog
computing uses local data (that is already present in the fog
nodes) and the data received from the IoT devices and per-
form near real-​time processing of data. Unlike MEC, since
the fog computing depends upon a single centralized server,
there is a huge possibility of service disruption to occur.
Cloudlets are a group of computing resources/​data centers
implemented on a smaller scale that are hosted in order
to improve the computing capabilities of mobile devices
within a geographical location/​proximity. It basically helps
to overcome the latency issues that are encountered by
mobile devices in a traditional wide area network (WAN) by
hosting the Cloud resources physically closer to the mobile
devices that require them. A Cloudlet is particularly different
from a public data center in that the Cloudlet is personally
managed by the users/​businesses that use it while the public
Cloud is managed solely by the Cloud service provider. As
the Cloudlet is locally maintained, it obviously depends
upon the LAN for connectivity. Cloudlets are thus used by a
smaller, specific, and closed group of users unlike the public
Cloud. It is basically used in resource-​intensive applications
that involve speech recognition, virtual reality, machine
learning, augmented reality, and language processing.
Micro Data Centers are small-​scale customized data centers
with all the required Cloud resources in terms of computing,
processing power, and storage that can be deployed closer
to the edge devices and in locations where a traditional data
center cannot be practically set up. It thereby reduces cost
and latency issues. While the fog computing is a form of
distributed architecture, micro data centers (MDCs) are a
type of data center design. MDCs typically help to build
the edge computer in a real-​world scenario. Businesses can
scale up progressively based on their changing demands.
MDCs facilitate easy deployment as the required equipment
are preinstalled by vendors on server stacks and shipped
as is. Upon receipt all that must be done is just an instal-
lation of any additional local components followed by
 I o T- C l o u d C o n v e r g e n c e 63

connection to an existing power supply. Once the MDCs are

up and running, service providers configure and monitor the
equipment using various tools.
Cloud of Things is a mechanism that makes use of the edge
devices’ resources such as advanced-​feature mobile phones
and other handheld devices to form a virtualized Cloud infra-
structure. Here the IoT devices themselves communicate
with each other and perform computations and processing
of real-​time data without depending upon the Cloud. For
example, a car can send traffic-​related updates and warnings
to other cars and suggest alternate routes without the need
for a user to feed in data.

3 . 4 . 3 Da ta Of f loadin g an d Compu t at io n s
With the recent increase in usage of IoT applications that collaborates
with the Cloud for many resource-​intensive computations, of huge
volumes of data collected at the edges are transferred to remote sites
for processing. Although offloading is important in order to perform
computations and storage at a larger scale, it incurs significant cost,
causes latency, and lowers the bandwidth due to the transportation of
data to and from the Cloud. Is also poses security and privacy threats
to the data in transit. A considerable amount of research is being
done in this area in order to suggest efficient and optimal offloading
techniques to improve the system performance and reliability.
In reference [13], the authors have proposed prediction-​ based
offloading schemes that predicts any available opportunities for
transferring data in future between mobile nodes in a terminal to ter-
minal (T2T) network based on the topology varying with time, node
mobility pattern, and the temporal contacts. The proposed system
has been verified to provide results with improved delivery ratio and
reduced latency overhead in comparison to other methods.
In reference [14], the authors have proposed a genetic algorithm
based adaptive offloading (GS-​OA) for efficient communication in
the IoT–​Cloud Infrastructure. The mechanism serves to improve the
response rate to requests with the help of the fitness process that is
distributed among the gateways and other devices in the infrastruc-
ture. This in turn leads to lesser delay and processing time in the
system, leading to better performance and improved throughput.
In reference [15], the authors have proposed a fog-​ based data
offloading mechanism to provide a reliable means of data collection
64 I o T- C l o u d C o n v e r g e n c e

and improved scalability in dense and dynamic environments. The

systems construct a multi-​tier data offloading protocol with the help of
the fog networking concept, thereby supporting a variety of data-​centric
IoT applications in the urban environments. The sensors present in the
heterogeneous network offload the data among themselves or to the
mobile gateways. The data drop-​off rates are monitored and minimized
based on the data pertaining to the mobility of humans and the urban
environment in which they operate upon. As a result, this mechanism
provides a lightweight and efficient method for data offloading.
In reference [16], the authors have attempted to solve the traffic
congestion problems encountered in IoT applications by proposing a
game-​based data offloading mechanism that leverages the Vickrey–​
Clarke–​Groves (VCG) mechanism and Rubinstein bargaining model.
The proposed scheme also addresses another aspect of data offloading,
which is incentivizing the access point owners. Here the mobile net-
work owners are responsible for properly incentivizing the access
point owners so that they will offload the traffic of non-​registered
users to the network as intended.
In reference [17], the authors have proposed an efficient mechanism
that saves time and energy in the process of selection of an appropriate
fog device (FD) for offloading. A module placement method by classi-
fication and regression tree algorithm is presented wherein decision on
the selection of the best FD is done based on the power consumption of
the fog device and other parameters such as authentication, confidenti-
ality, speed, availability, integrity, capacity, and cost.
In reference [18], the authors have presented a computational
offloading mechanism to address the challenges encountered in
offloading heterogeneous and huge volumes of mobile data to arbi-
trary resources. The mechanism in turn claims to improve the effi-
ciency, time, and energy consumption of mobile devices. In the
process, the system adopts the non-​dominated sorting genetic algo-
rithm III (NSGA-​III) to solve the problem of multi-​objective opti-
mization in task offloading in edge computing.
In reference [19], the authors introduce a framework that enables
device-​ to-​
device edge computing and network (D2D-​ ECN) for
data processing and computational offloading. This includes many
resource-​rich devices that participate in the network for offloading
and computation. However, such devices are prone to fall short of
battery life when involved in such high-​ efficiency computational
tasks. Hence an energy harvesting technology based on reinforcement
 I o T- C l o u d C o n v e r g e n c e 65

learning is incorporated to the D2D-​ECN framework. Also, an online

approach with the underlying principle of Lyapunov optimization is
used for efficient offloading and resource management in dynamic
environments.
In reference [20], the authors have proposed a nested game-​
based computational offloading for mobile Cloud IoT (MCIoT)
systems. The MCIoT extends the capabilities of mobile and portable
devices, enabling them to execute different mobile applications. The
Rubinstein approach is employed to determine the remote offloading
computation part and then a Cloud resource is dynamically assigned
for the computation. An optimal solution for this offloading process is
achieved by employing the nested game approach.

3 . 4 . 4 Of flo adin g Con side ration s an d

Cha lleng es
Offloading basically is a mechanism that helps in moving the
resource-​intensive and energy-​ consuming task to high-​ end
resourceful devices, thereby reducing delay, increasing bandwidth for
relevant communications, and energy conservation. Offloading can
also be done to store data that may not be required in near future
(archival data). However, to obtain an optimal throughput and per-
formance of applications there are certain considerations with respect
to implementing offloading in the IoT systems.

• As the process of offloading by itself incurs cost, offloading

cannot be done all the time depending upon only the users’
requirements. A decision on when (based on the current situ-
ation) the offloading can be done is very crucial. For example,
consider the activity of load balancing among multiple fog
nodes. Offloading the data does not help or is inefficient if
all the nodes are minimally loaded or extremely loaded (i.e.,
equal proportion). Offloading helps only if the load in multiple
servers/​nodes differ greatly (i.e., one in heavily loaded and the
other is sparingly loaded).
• Not all data/​ computations need to be offloaded. Certain
computations that can be accomplished with minimal resources
need not be transferred all the way to the servers, which will
only add additional overhead. What needs to be offloaded is an
important decision to be made.
66 I o T- C l o u d C o n v e r g e n c e

• Depending upon the context and type of computation to be

performed, a decision on where the data is to be offloaded must
be made carefully. Despite the Cloud providing all the required
services in an efficient way, the latency involved in transfer of
data makes it debatable. Hence, alternate solutions discussed in
Section 3.3.3 can be taken into consideration.
• A decision on how to offload is also very important in the IoT
systems. We can offload the tasks at different levels depending
upon the applications needs as follows:
i. System level offloading is a mechanism wherein the entire
the complete edge application along with the operating
system on which it was running is pushed to the server/​
Cloud (e.g., CloneCloud).
ii. Application-​level offloading is a mechanism wherein the
multiple applications are run as services on servers and the
user/​client can use these applications in their system by
accessing them via Web services (e.g., DAvinCi)
iii. Method-​level offloading is a mechanism where the appli-
cation/​program is broken down to separate modules/​pieces
and depending upon the need certain modules are run
remotely on servers (e.g., remote procedure calls (RPC)).

Following are a few challenges encountered in the process of

offloading.

Scalability and resource allocation: Allocation of the desirable

number of resources at the location of task execution (i.e., edge
or Cloud) in order to improve the efficiency is a challenge.
There may be extreme situations where resources are either
idle/​
under-​utilized due to lesser load or resources heavily
loaded demanding frequent offloading of workload. Under such
circumstances, scaling up and down of resources respectively
for effecting utilization remains a challenge.
High software quality: Offloading framework partitioning and
users’ varied inputs affect the system performance greatly.
High-​quality software is required to optimize the framework
portioning and improve user experience.
Scalability of applications: Scaling up the applications to ser-
vice requests coming in from multiple users without affecting
the offloading process is a challenge in maintaining the
performance.
 I o T- C l o u d C o n v e r g e n c e 67

Service level agreements (SLA): When critical IoT applications

that handle sensitive data perform offloading, care should be
taken that SLAs are not breached. Trust on the service pro-
vider and SLA monitoring are critical factors to be followed up
during offloading process.
Security and privacy of user data: In the process of offloading,
the threat of data being interpreted, tampered, and misused
increases when in transit. Ensuring a secure transmission
medium and a resilient framework for the security and privacy
of user data is of prime importance.
Integration and interoperability: With the confluence of a variety
of heterogeneous networks, platforms, operating systems, and
devices in the IoT, integrating all of them to perform tasks in
a harmonious manner is a great challenge. Different service
providers and IoT middleware might employ different protocols
and privacy and security mechanisms, making it difficult to
synchronize them and bring about an organized behavior in
applications.
Energy consumption: The offloading process by itself consumes
energy and bandwidth, thereby reducing system performance.
An informed decision must be made about whether to go ahead
with the offloading process or not.

3.5 DYNAMIC RESOURCE PROVISIONING

3 . 5 . 1 The Re source Provision in g A ct iv it y an d
Req u i rem ents f or IoT
Resource provisioning is a mechanism by which Cloud services
including storage, infrastructure, network, and computing capabil-
ities are made available to the applications/​users on a pay-​per-​use
basis via various Cloud computing vendors. The actors (software
agents, real-​world objects, and human beings) in the IoT architec-
ture collaborate with each other in real time and share information,
which helps individual networks to connect to the backbone, thereby
helping the IoT to be realized over a wide range of objects. The
IoT sensors are capable of monitoring and collecting information
from the dynamically changing environments. To enable efficient
functioning and processing of this data, effective resource allocation/​
provisioning is of prime importance in order to carry out the neces-
sary tasks. One of the major challenges in resource provisioning is
68 I o T- C l o u d C o n v e r g e n c e

the allocation of an apt amount of resources for performing the task

while still minimizing the cost and maximizing resource utilization.
Resource provisioning is categorized into three types, namely, static
resource provisioning, dynamic resource provisioning, and user
self-​provisioning.
Static resource provisioning refers to a mechanism wherein the
resources are allocated in advance by means of a contract between
the customer and service provider for the necessary services. The
service provider prepares and provides the resources beforehand and
charges the customers a fixed/​flat fee or on a monthly basis. This is
generally used for applications where the workload is predictable
and the demand for services remains unchanged for a long period
of time.
Dynamic resource provisioning is a mechanism wherein the
resources are allocated by the provider based on the changing needs
of applications. Here the virtual machine (VM) technology is used,
which helps underlying physical resource to be separated into mul-
tiple virtual devices, thereby facilitating multiple tasks to execute
simultaneously and reallocation to support the dynamic demands.
All such virtual resources are maintained in a pool and based on the
incoming workloads to the Cloud, the scheduler makes an informed
decision about the allocation and provisioning of VMs for task exe-
cution. The scheduler is run at periodic intervals to perform the
following activities:

• Predict the probable amount of inflow of workloads in future.

• Determine the amount of VMs to be provisioned from the Cloud.
• Monitoring the waiting time of jobs in the queue and deciding
on allocation of VMs to the deserving jobs.
• Monitoring the VMs and release of idle VMs.

In this case, the users/​customers are charged on a pay-​per-​use basis.

Cloud bursting is a term used when a hybrid Cloud is created using
the concept of dynamic provisioning.
User self-​provisioning is a mechanism wherein resources are avail-
able for purchase to the users via a Web form where they create an
account and pay for the resources using a credit card. The resources
are then made available with a few hours.
When considering an IoT scenario, the best fit resource provisioning
mechanism is the dynamic resource provisioning considering the
dynamically changing environment and needs of the application.
 I o T- C l o u d C o n v e r g e n c e 69

3.5.1.1 Resource Provisioning Requirements for an

IoT-​based Environment
The key features to be considered in the process of dynamic resource
provisioning for IoT environments are categorized as follows:

• Cost-​aware allocation refers to the overall price in terms of rev-

enue, resources, and profit. Maximizing the revenue and profit
and the cost-​effective allocation of resources to users (min-
imum expense) is important in an IoT environment.
• Efficiency-​aware allocation refers to the attempts made toward
improving the performance (number of tasks that are run
to completion) of the system by reducing the time taken to
respond to and execute a task/​workload. Also, an increase in
the execution speed (the amount of time the workload spends in
running to completion on an allocated resource) and bandwidth
(the amount of data that a particular connection can handle over
a specified period) with importance given to priority of tasks
(based on waiting time or type of workload) is important for an
efficiency-​aware system.
• Load-​balancing-​aware allocation refers to the ability to dis-
tribute the Cloud resources among multiple users. It also refers
to the capability to efficiently distribute the workload among
different resources across data centers without any disruption
to task execution.
• QoS-​aware allocation refers to the process of improving
user experience with respect to availability, throughput (total
number of tasks executed in a particular period), response time
(time taken to service/​attend to the request once it is out of the
waiting queue), security, and reliability (assurance to lowered
or no failure and guaranteed execution of task).
• Power-​aware refers to the ability to lower the level of energy
consumption across all levels in the IoT architecture. Energy is
basically required for resource preparation and execution of the
task in the IoT architecture.
• Context-​aware allocation means the ability system to collect
information from its environment dynamically and change its
behavior in a guided manner to enhance performance and user
experience.
• SLA-​based allocation means the ability of the system to ensure
that the SLAs (agreement between the service provider and
70 I o T- C l o u d C o n v e r g e n c e

users) are not violated, thereby providing fulfilled and satisfac-

tory services to the user.
• Utilization-​aware allocation refers to the ability to improve
the efficiency of the system by improving the utilization of
resources with reduced or no idle time, thereby reducing costs
incurred for the users.

3 . 5 . 2 Dyna mic Re source Provision i n g

In this section, we shall be exploring a brief overview of the existing
approaches employed in realizing dynamic resource provisioning. In
Ref. [21], the authors have proposed a priority-​based pre-​emption
policy that helps in addressing the problems of resource allocation in
a virtualized environment. As discussed, earlier virtualization helps in
dynamically allocating resources in order to ensure timely execution
of tasks and improved efficiency in the system. The underlying prin-
ciple of the proposed methodology is based upon the priority of the
job, which is decided based upon the deadline of the job. A job with
higher deadline (lower priority) should not postpone the execution
of a job with lower deadline (high priority). Allocation of VMs and
thereby the services before the deadline is the focus.
In Ref. [22] the authors have presented a solution based on
fog computing that provides a context-​aware framework that has
decision-​making modules dispersed across the IoT Cloud platform
and gateways. The solution presented makes decisions by analyzing
the data input from various sensors and the context of the surrounding
environment obtained from the fog computing resources. It also
manages the service provisioning activity based on the changes in
topology. Both the activities of service provisioning and decision
making are accomplished in one framework. Experimental results
have proven that the framework is resilient and responsive to changes
in topology.
In Ref. [23], the authors have proposed a mechanism that provides
a latency-​aware distributed resource provisioning solution for edge
network-​ based IoT applications. An algorithm that supports the
distributed allocation of resources in order to support flawless integra-
tion and deployment of various applications in the IoT infrastructure
is also presented. The algorithm performs a mapping of the applica-
tion to the edge network/​device and supports dynamic migration of
the application fragments abiding by the SLA.
 I o T- C l o u d C o n v e r g e n c e 71

In Ref. [24], the authors introduce a QoS provisioning framework

(QoPF), which is dynamic in nature and applicable to service-​oriented
IoT. The framework uses the backtracking search optimization algo-
rithm (BSOA). The proposed framework maximizes the IoT appli-
cation layer’s composite service quality by balancing the service’s
reliability and cost incurred for the computation time.
In Ref. [25], the authors have addressed the issues related to user’s
privacy that arises during the collection and transport of user data
in the IoT environment. The European Union has made it illegal for
the IoT devices and other sensors to collect people’s data without
obtaining consent from the user. To ensure such user privacy, a data
filtering mechanism has been introduced at the edge devices/​net-
work. Application of privacy rules to the data at the edge of the net-
work will enable restricted transport of data to the Cloud. However,
this additional processing can introduce some unexpected delay and
bottlenecks in the system. Hence as a countermeasure the authors
have introduced a hybrid data processing technique to address the
issues of privacy and resource provisioning.
Fog computing being more suitable to address the recent and
upcoming demands in IoT applications has become an integral part of
the IoT architecture. With this in mind, the authors in reference [26]
have presented a container-​based virtualized networked architecture.
The architecture finds its place in the middleware layer and works
by taking advantage of the native Container Engines, thereby enab-
ling the real-​time up scaling and down scaling of the virtual resources
such as computing and networking resources. The virtualized com-
puting and networking resources are dynamically managed by a less
complex penalty-​aware-​bin packing heuristic.
In reference [27], the authors have introduced a QoS-​aware frame-
work for dynamic resource provisioning in fog computing. The
framework addresses the dynamic deployment and release of appli-
cation services in fog nodes, thereby lowering the cost and latency
while still maintaining the QoS requirements. Two highly efficient
greedy algorithms have been used and the framework proves to be
efficient even with very minimal IoT node information and makes no
assumptions on the network.
In reference [28], the authors have proposed a model-​based ser-
vice provisioning mechanism that is multidimensional and multilevel
in nature. The platform renders the potential of large-​scale heteroge-
neous resources as lightweight services. An integrated message space
72 I o T- C l o u d C o n v e r g e n c e

has been introduced to aid the sharing of sensory information/​data in

a distributed IoT environment. The platform also encourages sharing
and reuse of resources.

3.6 SECURITY ASPECTS IN IOT CLOUD

CONVERGENCE
Form the above discussions, it is evident that Cloud computing has
indeed evolved with new features in order to provide for the needs of
IoT. These features include advanced processing and decision-​making
capabilities, virtualization mechanisms, and analytical tools. Most of
these features have been discussed in the previous sections. However,
all these advanced capabilities come with various risks and threats
to the security of the IoT system. This chapter aims at discussing the
various security challenges that accompany IoT–​Cloud convergence.

3 . 6 . 1 ENIS A’s Categorized Se curit y C h allen ges

i n Io T Clo ud Con ve rge n ce
The Cloud being an integral part of the IoT has many security-​related
issues due to the numerous devices with diverse specifications and
protocols that are connected to it. The European Union Network
and Information Security Agency (ENISA) categorizes the security
aspects that come with the IoT Cloud convergence based upon the IoT
reference model as follows:

• Connectivity refers to the interfaces (communications and

interactions) among the components such as the endpoints
(devices, sensors, actuators etc.,), gateways and the Cloud.
According to ENISA, the two major contributors toward the
connectivity-​ related security issues are the heterogeneous
protocols that are used for communication and the insecure flow
of data from the edge devices into the IoT system. The diverse
range of applications where IoT is used requires the incorpor-
ation of multiple heterogenous hardware, protocols, operating
systems, network connection, and so on in the IoT system. In
certain applications, there is a need to integrate legacy systems
with new devices, which in turn could lead to unprecedented
risks in the security of the system.
Another aspect of the IoT system includes the processing of
data, which in an IoT system is done at the Cloud or at the
 I o T- C l o u d C o n v e r g e n c e 73

edge devices as discussed in Section 3.3.2. Edge computing

enhances the processing capabilities of edge devices, thereby
providing major advantages such as reducing the latency
involved in transfer of data to the Cloud and back to the devices,
reducing costs, bringing near real-​time operations closer to the
data sources, and many more. However, the restricted capacities
of the edge devices in terms of storage and computing power
make it difficult to incorporate robust security mechanisms and
algorithms, thereby making them vulnerable to security threats.
Faulty implementations can lead to a compromise on the integ-
rity, confidentiality, and the privacy of the data.
• Analysis includes data processing, filtering, and aggregation
that comes into the various levels of the IoT system, which is
collected by the endpoint devices. ENISA has identified that the
real-​time processing of data at the edge devices and the decen‑
tralization of Cloud are two aspects that majorly contribute
toward compromising the security of the IoT system. The very
nature of the environment in which the edge devices operate
(common places where security implementations are nil) makes
it easier for attackers to gain access and manipulate the data.
Also, security functionalities such as monitoring and tracking
which is performed at the Cloud becomes difficult to imple-
ment and manage at the edge. Additionally, Cloud decentral-
ization that enables better management and processing, thereby
reducing the overhead on a single remote Cloud, also comes
with its challenges. The edge devices that lack elasticity unlike
the Cloud are more prone to denial of service (DoS) attack
following which the coordination of security mechanisms such
as security patches and software also increases.
• Integration represents Cloud APIs and other such features that
facilitates the real-​time bidirectional data flow in the system.
According to ENISA, security can be greatly influenced by
the kind of environment/​vertical that the Cloud serves. For
example, an IoT system that functions at a critical operations
environment such as a smart hospital, security is greatly
influenced based on the complicated infrastructure and the
management policies. Security also depends upon the know‑
ledge and implementation expertise levels of the developers.
Developers should be completely aware of the primary elem-
ents of security in the smart devices such as message authen-
tication and encryption, authentication and identification of
74 I o T- C l o u d C o n v e r g e n c e

things, message integrity, secure booting, key management, and

necessary storage implementations. Also, it becomes difficult to
lay out a generalized and standard hardware/​software develop-
ment life cycle for the IoT systems due to the involvement of
varied degree of hardware, protocols, and network connectivity
mechanisms.

3 . 6 . 2 Io T S ec urity Design s Base d o n E dge

Co m p uti ng
Edge computing has been discussed previously in many sections.
Having touched upon the basics of edge computing and the advantages
it has to offer to the IoT system, it is evident that edge computing
offers a great deal in optimizing the system performance. However,
the devices are also vulnerable to security threats like any other
system and hence arises the need to implement security solutions on
these edge devices. Although edge device security implementations
are at its infancy, considerable amount of research has been conducted
in the area. This section aims at discussing few of the notable security
solutions for edge devices.
A quick look at the edge-​centric IoT architecture will provide a
clear understanding to the readers on the design and deployment of
the security solutions in the appropriate layers. Figure 3.2 represents

Figure 3.2 A high-​level edge-​centric IoT architecture.

 I o T- C l o u d C o n v e r g e n c e 75

a high-​level edge-​centric IoT architecture. At the lowest level are the

IoT end devices (thermostat, electric bulb, smart watches, displays
etc.,) that act as sensors and collect information from the surrounding
environment. The next layer comprises the edge devices that are
equipped to store and process considerable amount of data (Raspberry
Pi, autonomous cars, smartphones, personal computers etc.). These
edge devices assist in performing computations and other activities
for real-​time scenarios. The final layer is the Cloud wherein resource-​
intensive operations, decision-​making, and more are carried out.
The edge layer also acts as the best candidate to deploy/​implement
the security solutions for various reasons. First, as mentioned earlier
the edge devices being equipped with better storage and processing
capabilities in comparison to the end devices, thereby making it feas-
ible to implement complex security operations such as encryptions,
attribute-​based access control schemes, and so on. Second, real-​time
security needs can be catered to by the edge devices due to the prox-
imity to the end devices. Third, as mentioned earlier, the edge device
collects and stores the information collected by end devices for further
processing. This makes it a suitable place to support security-​related
decisions as the collected data can supply the required information.
For instance, more data helps in detecting intrusions more efficiently.
Fourth, considering the widespread presence of end devices, setting
up firewalls across the end devices is very intimidating. However,
firewalls can be deployed at the edge layer, which in turn can filter
any malicious requests/​attacks. The highly mobile end devices can
be monitored and tracked by the edge devices efficiently. Also, as
the end devices constantly send information to the appropriate edge
devices, trust can be effectively established between the edge and end
devices. This eliminates the overhead of establishing trust between
the end devices themselves. Lastly, the edge devices are most often
connected to the Cloud via high-​speed links, which makes it easier for
edge devices to rely on the Cloud for any resource-​intensive security
computations and support. With the edge-​centric IoT architecture in
mind, we move on to discussing a few and most relevant edge-​based
security implementations for IoT.
User-​centric edge-​based IoT security architecture: The end users
of the IoT applications play a major role in making the IoT applications
operational. They gain access to many IoT resources via terminals
such as the smartphones, personal computers, smart watches, and so
on. The resources are made available to the users in a more convenient
and pervasive manner. However, users who access these resources
76 I o T - Cl o u d C o n v e r ge n c e

cannot be expected to access the resources via secure and trusted

terminals. Also, not all users possess the knowledge and awareness
of various security measures. Hence, management of security must
be incorporated at a more reliable layer, which is the edge layer. This
includes setting up measures to manage personal security of users and
positioning virtual security systems at the network edge.
Device-​centric edge-​based IoT security design: Devices in the IoT
architecture collect information from the surrounding environment
and act as actuators that help in driving the decision-​making process
that control the physical environment. In the device-​centric edge-​based
IoT security design, the security controls are offloaded from the end
devices to the edge layer. It provides customized security solutions
for each individual device based on data that the device is sensing, the
effect of actuating a task as a response, the availability of resources,
and the security requirements of a collection of end devices. EdgeSec
[29] and ReSIoT [30] are a few designs that represent the device-​
centric security architecture.
End-​to-​end IoT security: A security system that covers all the
components of the IoT system and the communications between them
is ideal for realizing a reliable application. However, the heterogen-
eity of the participating devices, Cloud, and the protocols make it
difficult to attain the goal. Researchers have identified the edge layer
as a suitable candidate to house the security controls as it operates
as an intermediary between the heterogenous devices and the Cloud.
A secure middleware setup at the edge layer helps in establishing a
secure end-​to-​end communication among the participating devices
[31]. The middleware manages security functions such as the encryp-
tion algorithms, authentication mechanisms, MAC algorithms, and
secure sessions for devices.
Firewalls implementation at the edge layer: Heavy resource
demanding applications like firewalls cannot be run on resource-​
constrained devices. Also, the huge number of IoT end devices that
are connected ubiquitously to the IoT makes it extremely costly to
deploy and manage the firewall on each individual device. Hence,
installing firewalls in the edge layer is an ideal choice due to the
following reasons.

• Firewall updates can be more achievable due to the existence of

one centralized conceptual firewall.
• As in most cases where an edge device manages a subsystem
of the IoT application, the firewall can be deployed in the edge
device that manages the entire subsystem.
 I o T- C l o u d C o n v e r g e n c e 77

• User mobility can be traced with the edge device monitoring the
user movement and credentials of the users and devices.

Firewall policies that are defined by the IoT application are translated
into flow policies. These flow policies are revisited to rule out any
conflicts after which the policies become a set of firewall rules
that are installed at the edge. All incoming and outgoing traffic are
then scrutinized against these set of rules. Examples of firewall-​
based security at the edge layer include FLOWGAURD [32] and
a distributed firewall architecture [33], which use software-​defined
networks (SDN) and virtual network function (VNF) respectively.
Intrusion detection system at the edge layer: In 2016, a successful
distributed denial of service (DDoS) attack was launched by attackers
by compromising a great number of IoT end devices. This disrupted
internet services in a wide range of areas. The presence of an effi-
cient intrusion detection system would have helped address the issue,
thereby minimizing the loss incurred. Designing an intrusion detec-
tion system at the edge layer would be more efficient considering the
availability of huge data/​information that is transferred from mul-
tiple end devices/​sources. This data can then be run against machine
learning algorithms to identify any intrusions and adapt to changes
in attack patterns. A typical design of the intrusion detection at the
edge layer as shown in reference [34] consists of a distributed traffic
monitoring system that collects the network traffic in real time. The
data is then run against the intrusion detection algorithm locally at
the edge device. The collective data from multiple edge devices are
also collaboratively monitored for any intrusion. Finally, the network
controllers installed at the edge devices impose the detection outcome
for further action.
Edge-​based authentication and authorization: Authentication and
authorization play a major role in security that helps prevent many
attacks such as unauthorized access and the DDoS attacks. However,
the huge number of devices and heterogenous nature of platforms and
protocols make it difficult to establish a strong security mechanism.
Also, standard security aspects such as the digital signatures cannot
be applied to the IoT ecosystem. These problems can be resolved
by the edge layer that acts as a middleman between the end devices
and the Cloud. The authentication process can be broken down into
multiple segments with the edge being the pivot. The first segment
involves the authentication between the end device and the edge, and
the second segment includes the authentication between the edge and
the Cloud/​user/​other end devices. Authentication protocols can be
78 I o T- C l o u d C o n v e r g e n c e

tailored depending upon the characteristics of the participating entities

in the segment. The edge also acts as a proxy for the end devices in
the authentication and authorization process. With the edge devices
having the necessary resources to support the computations and pro-
cessing power required for the algorithms the end-​to-​end authentica-
tion and authorization is made possible in the IoT ecosystem.
Edge-​based privacy-​preserving techniques: With numerous users
connected over the internet and in turn the IoT applications, they share
information that is required for basic communication and carrying
out task in the system. Despite the IoT applications being an inevit-
able part of everyone’s lives, users require strict privacy-​preserving
mechanisms in place. With edge layer being rich in information/​data
collected by the end devices, it is ideal to implement various privacy-​
preserving algorithms at this layer. When queries are presented via an
IoT application, the edge layer processes the information against the
defined goals and algorithms in order to provide the applications with
data that is privacy preserved.

3.7 IOT–​C LOUD CONVERGENCE: TEST BEDS AND

TECHNOLOGIES
This section introduces us to a few testbeds that is a part of the FIESTA-​
IoT project [35]. The project aids in experimenting new approaches
toward the development and deployment of IoT applications by lever-
aging the possibility of interconnection of varied test beds and platforms.
It also aims at making the process of sharing data between various
testbeds and easy and seamless task to enable research and development.

3 . 7 . 1 Over view of Testbeds an d Pl at f o r ms

3.7.1.1 SmartSantander Testbed

The SmartSantander is an experimental testbed. It is used for
experimenting various architectures, future technologies, and services
targeted for Internet of Things and applications. It is usually done in
the context of a city. SmartSantander is platform scalable with new
and diverse variety of sensors. The architecture of SmartSantander
testbed is divided into three tiers.

IoT tier: This tier is responsible for sensing the data of interest
like temperature, chemical composition of atmosphere,
 I o T- C l o u d C o n v e r g e n c e 79

sound, light, presence of a vehicle parked, temperature, and

humidity of soil. Sensors are built with networking capability.
The network module built in them is called repeaters. Some
of the sensors are standalone, in that case they use wireless
technology to communicate with the repeaters. Few of these
devices are powered by batteries due to the location in which
they are placed. Repeaters are located at high-​ rise places
like streetlights, public buildings, and relay the information
captured via sensors. Digimesh also known as IEEE 802.15.4
is the preferred mode of communication between repeaters and
IoT nodes. For experimentation, a native Digimesh interface is
available in the repeaters.
Gateway tier: Gateway receives all information that is sent from
IoT nodes and repeaters. The received information is forwarded
to the top layers of SmarSantander intelligently and efficiently.
It additionally has tons of networking tools and control and
manage the network.
Platform tier: This topmost layer can ingest huge volume of
data from the gateway. It has various interfaces to handle the
incoming data. Platform management tools are also available
here. IoT device registration repository is available in this
layer. Information received from the sensors is also stored in
this layer.

3.7.1.2 UNIS Testbed

UNIS Testbed has three tiers:

Server tier: The server tier is the entry point for users to access
the test bed and perform operations on it. All the back-​end
components and its functionalities are hosted in the server tier.
Gateway tier: The gateway tier is the backbone of the test infra-
structure. It connects the server tier and the IoT tier via WiFi
or Ethernet. Gateway tier is also known as an embedded
gateway (GW).
IoT tier: As the name suggests, IoT tier consists of IoT sensor nodes.
It also comprises consumer electronics such as smartphones
and smart displays. IoT nodes can be connected in two ways.
The first way is wireless communication that includes WiFi and
Bluetooth. Second, a USB connection, which is wired, is used
for management of gateway.
80 I o T- C l o u d C o n v e r g e n c e

3.7.1.3 KETI Testbed

The KETI testbed built for indoors. It is an experimental test bed.
All indoor parameters such as human presence, humidity, indoor tem-
perature, and indoor luminance is monitored. Armed with this data,
an energy management system is built for the maintaining optimum
operational level of the building. KETI testbed is divided into three
tiers. They are server platform known as Mobius, device platform
known as &Cube, and final IoT device tier.

Mobius (server platform): The Mobius platform complies with

oneM2M standards. It is an open IoT service platform. A large
number of IoT devices are indexed so that devices of interest
can be easily discovered. RESTFul APIs are the preferred way
of data transfer and it complies with oneM2M standards.
&Cube (the device and gateway platform): It is designed on
JRE (Java Runtime Environment) to serve as the oneM2M
standards-​defined software platform for its home gateway and
Cloud platform. It could be ported anywhere to gateways or IoT
devices that can run a Java Virtual Machine (JVM).
IoT devices: IoT devices collect indoor sensory data. They transmit
the data to the gateways using Zigbee wireless mesh network.
Gateways are Raspberry Pis and installed with &Cube, which
is a software-​based gateway. JSON or XML is used to transfer
the data to the server platform Mobius.

3.7.1.4 Com4Innov Testbed

Com4Innov is an actual life-​ sized laboratory. Its environment is
already set up and in use. Com4Innov test bed is used to analyze
wireless technology and its corresponding services before the tech-
nology could be ready for the market. Few examples are network
applications and mobile services of the future, next gen communi-
cation objects, and smart devices for IoT are accompanied by data
handling and compute facilities.
FIESTA-​IoT project drives Com4Innov via 5G communication
technology. Huge amounts of data are generated due to 4G/​LTE net-
work interconnection with IoT infrastructure. Com4Innov data center
can host several types of services, for example Mobius from KETI
testbed and OpenIoT.

Networking technologies: Com4Innov has a 4G/​LTE core and an

access network. A highly noncentralized architecture of sensors
 I o T- C l o u d C o n v e r g e n c e 81

with methodologies and tools for monitoring and measuring is

used in the Com4Innov test bed.
Applications and services: The applications that are available
in Com4Innov test bed are IP Multimedia Core Network
Subsystem and Applications servers, Multimedia Telephony
Presence Management, and Instant Messaging. M2M Cloud
provides device to gateway and Cloud solutions for federation
and data and object management. M2M Cloud also provides
tools and simulators for sensors for data traffic along with
remote SIM card solutions.
Terminals: At the terminals, a plethora of 4G/​LTE terminals
as well as prototypes, SIM cards to access the 4G/​ LTE/​
IMS services, plethora of machine-​to-​machine sensors, and
SIM cards for machine-​to-​machine tests on real network are
provided.
4G/​IMS Testbed is a IPV6/​IPV4 supported network with Wi-​Fi
calling and SMS over IMS capability. It also supports tracing
and logging. Radio channels are available with frequency div-
ision duplex (FDD)/​time division duplex (TDD) support. User
simulation supported on radio access network (RAN). 4G user
equipment like phone, tablets, and USB dongles are used.
Gateways and sensors: Sixty sensors are available to build
machine-​to-​machine test case or proof of concept. Here are a
few of the supported sensors. Thermal current flow sensor, sun-
light measurement, rain measurement, and wind speed measure-
ment sensors, crack measurement in concrete sensors, concrete
temperature measurement sensor, carbon dioxide detection and
noise pollution detection sensors. The gateway supported are
2G/​3G/​4G/​Internet.

REFERENCES
1. https://​csrc.nist.gov/​publications/​detail/​sp/​800-​145/​final.
2. Nancy Ambritta P., Poonam N. Railkar, and Parikshit N. Mahalle.
Proposed Identity and Access Management in Future Internet
(IAMFI): A Behavioral Modeling Approach, Journal of ICT, 2(1),
1–​36, 2014.
3. Lu Hou, Shaohang Zhao, Xiong Xiong, Kan Zheng, Periklis
Chatzimisios, M. Shamim Hossain, and Wei Xiang. Internet of Things
Cloud: Architecture and Implementation, IEEE Communications
Magazine —​Communications Standards Supplement, 54, 32–​39,
2016.
82 I o T- C l o u d C o n v e r g e n c e

4. Hany F. Atlam, Ahmed Alenezi, Abdulrahman Alharthi, Robert

J. Walters, and Gary B. Wills. Integration of Cloud Computing
with Internet of Things: Challenges and Open Issues, IEEE
International Conference on iThings-​ GreenCom-​ CPSCom-​
SmartData, 17, 670–​675, 2017.
5. Ademir F. da Silva, Ricardo L. Ohta, Marcelo N. dos Santos, and
Alecio P.D. Binotto. A Cloud-​based Architecture for the Internet
of Things targeting Industrial Devices Remote Monitoring and
Control, IFAC-​PapersOnLine, 49(30), 108–​113, 2016.
6. Dimitrios Kelaidonis, Panagiotis Vlacheas, Vera Stavroulaki,
Stylianos Georgoulas, Klaus Moessner, Yuichi Hashi, Kazuo
Hashimoto, Yutaka Miyake, Keiji Yamada, and Panagiotis
Demestichas, Cloud Internet of Things Framework for Enabling
Services in Smart Cities, In: Angelakis V., Tragos E., Pöhls
H., Kapovits A., Bassi A. (eds) Designing, Developing, and
Facilitating Smart Cities, pp. 163–​191, Springer, Cham, 2017.
7. Yu Liu, Kahin Akram Hassan, Magnus Karlsson, Zhibo Pang, and
Shaofang Gong, A Data-​Centric Internet of Things Framework
Based on Azure Cloud, IEEE, 7, 53839–​53858, 2019.
8. Laura Belli, Simone Cirani, Luca Davoli, Lorenzo Melegari,
MàriusMónton, and Marco Picone, An Open-​ Source Cloud
Architecture for Big Stream IoT Applications, In: Interoperability
and Open-​Source Solutions for the Internet of Things, pp.73–​88,
Springer, Cham, 2015.
9. Tian Wang, Guangxue Zhang, Anfeng Liu, Md ZakirulAlam
Bhuiyan, and Qun Jin. A Secure IoT Service Architecture with
an Efficient Balance Dynamics Based on Cloud and Edge
Computing’, IEEE, 6, 4831–​4843, 2018.
10. Asad Javed, Jerémy Robert, Keijo Heljanko, and Kary Främling.
IoTEF: A Federated Edge-​Cloud Architecture for Fault-​Tolerant
IoT Applications, Journal of Grid Computing, 18, 57–​80, 2020.
11. Lorenzo Carnevale, Antonio Celesti, Antonino Galletta, Schahram
Dustdar, Massimo Villari. From the Cloud to Edge and IoT: a Smart
Orchestration Architecture for Enabling Osmotic Computing,
IEEE 18, 419–​424, 2018.
12. Aliza Jamal, Farhan Ahmed Siddiqui, Adnan A. Siddiqui, Nadeem
Mahmood, Muhammad Saeed, and Syed Asim Ali. Algorithms
and Techniques for Computation Offloading in Edge Enabled
Cloud of Things (ECoT) –​A Primer, International Journal of
Computer Science and Network Security, 19(6), 1–​11, 2019.
13. Ankan Ghosh, Osman Khalid, Rao N. B. Rais, Amjad Rehman,
Saif U. R. Malik, and Imran A. Khan. Data offloading in IoT
Environments: Modeling, Analysis, and Verificatin. EURASIP
Journal on Wireless Communications and Networking, 53, 2019.
 I o T- C l o u d C o n v e r g e n c e 83

14. Azham Hussain, S. V. Manikanthan, T. Padmapriya, and

Mahendran Nagalingam, Genetic Algorithm Based Adaptive
Offloading for Improving IoT Device Communication Efficiency,
Wireless Networks 26, 2329–​2338, 2020 .
15. P. Kortoçi, L. Zheng, C. Joe-​ wong, M. D. Francesco, and
M. Chiang. Fog-​ based Data Offloading in Urban IoT
Scenarios, IEEE INFOCOM 2019 –​IEEE Conference on
Computer Communications, pp. 784–​792, 2019.
16. Youngjae Park, Sungwook Kim. Game-​Based Data Offloading
Scheme FORIOT System Traffic Congestion Problems, EURASIP
Journal on Wireless Communications and Networking, 192, 2015.
17. Rahbari, D., Nickray, M. Task offloading in mobile fog computing
by classification and regression tree. Peer-​to-​Peer Networking and
Applications 13, 104–​122, 2020.
18. Xiaolong Xu, Qingxiang Liu, Yun Luo, Kai Peng, Xuyun Zhang,
Shunmei Meng, and Lianyong Qi. A computation offloading
method over big data for IoT-​ enabled cloud-​ edge computing,
Future Generation Computer Systems, 95, 522–​533, 2019.
19. Qiao G., Leng S., and Zhang, Y. Online Learning and Optimization
for Computation Offloading in D2D Edge Computing and
Networks, Mobile Networks and Applications, 2019.
20. Kim, S. Nested game-​ based computation offloading scheme
for Mobile Cloud IoT systems. EURASIP Journal on Wireless
Communications and Networking, 229, 2015.
21. A.T. Saraswathi, Y.R.A. Kalaashri, and S. Padmavathi, Dynamic
Resource Allocation Scheme in Cloud Computing, In: Procedia
Computer Science, vol. 47, pp. 30–​36, 2015.
22. Pešić S., Tošić M., Iković O., Ivanović M., Radovanović M., and
Bošković D. Context Aware Resource and Service Provisioning
Management in Fog Computing Systems. In: Ivanović M., Bădică
C., Dix J., Jovanović Z., Malgeri M., Savić M. (eds) Intelligent
Distributed Computing XI. IDC 2017. Studies in Computational
Intelligence, vol. 737. Springer, Cham, 2018.
23. Avasalcai C., Dustdar S. Latency-​Aware Distributed Resource
Provisioning for Deploying IoT Applications at the Edge of the
Network. In: Arai K., Bhatia R. (eds) Advances in Information
and Communication. FICC 2019. Lecture Notes in Networks and
Systems, vol. 69. Springer, Cham, 2020.
24. Badawy M.M., Ali Z.H., and Ali H.A. QoS provisioning frame-
work for service-​ oriented internet of things (IoT). Cluster
Computing 23, 575–​591, 2020.
25. Kirkham T. et al. (2016) Privacy Aware on-​Demand Resource
Provisioning for IoT Data Processing. In: Mandler B. et al. (eds)
Internet of Things. IoT Infrastructures. IoT360 2015. Lecture
84 I o T- C l o u d C o n v e r g e n c e

Notes of the Institute for Computer Sciences, Social Informatics

and Telecommunications Engineering, vol. 170. Springer, Cham.
26. Vinueza Naranjo P.G., Baccarelli E., and Scarpiniti M. Design and
energy-​efficient resource management of virtualized networked
Fog architectures for the real-​time support of IoT applications,
Journal of Supercomputing 74, 2470–​2507, 2018.
27. Yousefpour A. et al. FOGPLAN: A Lightweight QoS-​ Aware
Dynamic Fog Service Provisioning Framework, IEEE Internet of
Things Journal, 6(3), 5080–​5096, 2019.
28. Zhao S., Yu L., and Cheng B. An Event-​ Driven Service
Provisioning Mechanism for IoT (Internet of Things) System
Interaction, IEEE, 4, 5038–​5051, 2016.
29. Sha K., Errabelly R., Wei W., Yang T.A., and Wang
Z. EdgeSec: Design of an Edge Layer Security Service to Enhance
IoT Security 2017 IEEE 1st International Conference on Fog and
Edge Computing (ICFEC), Madrid, Spain, pp. 81–​88, 2017.
30. Hsu R., Lee J., Quek T.Q.S., and Chen J. Reconfigurable
Security: Edge-​ Computing-​ Based Framework for IoT, IEEE
Network, 32(5), 92–​99, 2018.
31. Mukherjee B., Neupane R.L., and Calyam P. End-​to-​End IoT
Security Middleware for Cloud-​Fog Communication, IEEE 4th
International Conference on Cyber Security and Cloud Computing
(CSCloud), 2017, pp. 151–​156.
32. Hongxin Hu, Wonkyu Han, Gail-​Joon Ahn, and Ziming Zhao.
FLOWGUARD: building robust firewalls for software-​ defined
networks. In Proceedings of the third workshop on hot topics
in software defined networking (HotSDN '14). Association for
Computing Machinery, New York, NY, USA, 97–​102, 2014.
33. T. Markham and C. Payne. Security at the network edge: a
distributed firewall architecture, Proceedings DARPA Information
Survivability Conference and Exposition II. DISCEX'01,
Anaheim, CA, USA, pp. 279–​286, vol. 1, 2001.
34. Kewei Sha, T. Andrew Yang, Wei Wei, and Sadegh Davari.
A survey of edge computing-​ based designs for IoT security,
Digital Communications and Networks, 6(2), 195–​202, 2020.
35. FIESTA-​IoT, www.fiesta-​iot.edu
 4

SMA RT CO M PUT I N G
O VER IO T –​CLO UD

4.1 INTRODUCTION
Smart computing in general refers to the mechanism of empowering
devices and utilities that we use in our day-​to-​day lives with com-
puting capabilities. On similar lines, smart computing over the
IoT–​Cloud refers to confluence of hardware, software, and network
technologies that empower the IoT–​Cloud application with real-​time
awareness of the environment and enhanced analytics that can assist
humans in better decision making and optimizations, thereby driving
the business to success.
This chapter discusses smart computing technologies that are
involved in driving the IoT–​Cloud applications to success. Section
4.1 provides an overview of big data analytics and cognitive com-
puting from an IoT–​Cloud perspective. Section 4.2 provides a brief
explanation of a few deep learning approaches. Section 4.3 discusses
the techniques, algorithms, and methods and Section 4.4 includes a
few case studies pertaining to smart computing over IoT–​Cloud.

4.2 BIG DATA ANALYTICS AND COGNITIVE

COMPUTING
From the earlier chapters, we now have a clear understanding about
IoT–​Cloud. With the innumerable devices that are connected to each
other and the Cloud through the Internet, the amount of data that these
devices collect or generate is immeasurable and huge. The size of this
data can be in petabytes of gigabytes. Additionally, given the hetero-
geneity of these devices, platforms, and environment of operation,
the data that these devices collect are present in different formats.
They come in a variety of forms such as structured (an organized

85
86 Smart Computing over IoT–Cloud

format), unstructured, and semistructured. This huge volume of data

in different formats is referred to as big data and analysis of this
data in order to generate inferences and solutions is called as big
data analytics. However, analysis of such data in IoT applications is a
huge challenge due to the their size and heterogeneity.
Cognitive computing is a mechanism used in solving problems
that are complex and may have a certain degree of uncertainty in
arriving at suitable answers. It is a self-​learning system that mimics
the human brain/​thinking with the help of computerized models. It
is a confluence of several underlying technologies such as natural
language processing (NLP), pattern recognition, data mining, senti-
ment analysis, machine learning, neural networks, and deep learning,
which we will discuss in detail in this section. The confluence of big
data analytics and cognitive computing will help unleash the greater
potential of IoT–​Cloud applications. This section will discuss about
the cognitive computing capabilities, provide a detailed descrip-
tion of the various technologies used and how cognitive computing
empowers analytics.

4 . 2 . 1 Co g nitive Computin g Capabilit ies

In today’s IoT-​ connected world, the number of data generators
(human, machines, and other business applications) are multiplying
rapidly, consequently generating huge volumes of structured and
unstructured data. This has led to the development of an economy that
is data-​driven. This has indirectly forced many businesses to venture
into cognitive computing and enhanced analytics in order to stay rele-
vant and competent. Cognitive computing capabilities are immense
and multidisciplinary.
Cognitive computing can offer improved data analysis. For
instance, the health care industry assimilates data from various
sources such as journals, medical records, diagnostic tools, and other
documents. All these data provide evidence and help make informed
decisions and recommendation related to the treatment that can be
provided to patients. Here is where cognitive computing comes in
handy by performing quick and reliable analysis of the data and
presenting it to the physicians, surgeons, or medical professionals.
Cognitive computing can lead to improved customer satisfaction
levels. For instance, the Hilton group [1], which is a hospitality and
travel business, has employed a robot, Connie (Watson enabled) that
provides customers with precise, relevant, and accurate information
 Smart Computing over IoT–Cloud 87

on various topics related to travel and accommodation. It also provides

information on fine dining, amenities offered at hotels, and places to
visit thus making the customers have a smart, easy, and enjoyable
travel experience.
Cognitive computing can simplify complex processes into simpler
and efficient processes. In the case of Swiss Re [1], an insurance com-
pany, the application of cognitive computing has made the process
of identifying patterns simpler and efficient, thereby enabling real-​
time problem solving for more efficient responses. It has employed
the IBM Watson technology to perform analysis of huge volumes of
structured and unstructured data pertaining to the risk of exposure of
sensitive information. Based on the analysis, measures were adopted
to put efficient risk management tools in place and improve the prod-
uctivity of the business.
Employing cognitive computing can help improve businesses by
dealing with financial services for purchases. For instance, customers
wanting to buy automobiles can have varying levels of financial
packages and services. The agents can employ cognitive technologies
with the consent of customers in order to surf through their data such
as credit score, and other important financial documents in order to
suggest offers that would benefit both the buyer and seller.
Cognitive computing can be employed for identifying safety
concerns in a product earlier in the lifecycle, thereby helping to
reduce costs that might be incurred in a recall after completion. It also
helps in upholding reputations of big organizations by identifying
shortcomings at an earlier stage. Also, the delays in time-​to-​market
that might occur if a product fails are also taken care of with early
detection.
The banking sector has reaped great benefits by employing cogni-
tive computing. In particular, fraud detection has an improved edge
where the earlier rule-​based detection mechanism (based on standard
questions) has been expanded to mechanisms that analyze banking
behavior of customers (spending habits), mechanisms that predict
future expenditure or buying based on previous patterns, and many
such mechanisms. This helps in detection of any anomaly in the
regular patterns, leading to card blockage and alert sent to the legit-
imate customer.
Cognitive computing over IoT can enable products to make inde‑
pendent and instantaneous decisions in businesses without human
interference. Fact-​ based solutions can be provided proactively to
drive the entire business process right from engaging in relevant and
88 Smart Computing over IoT–Cloud

meaningful conversations with customers to the manufacturing and

maintenance of tools and equipment.
Cognitive computing must possess the following features in order
to realize the above-​mentioned capabilities.

• Adaptive: Cognitive computing must be able to keep up with

the dynamically changing data, goals and requirements by
learning, and updating constantly.
• Interactive: Cognitive computing should provide flexibility
and ease by allowing users to communicate just the way they
would in a real-​world human-​to-​human interaction using voice,
gestures, and natural languages.
• Iterative and stateful: It should possess the capability of collecting
relevant information by asking suitable questions from the user
in the event where enough information and requirements are not
available in order to describe the problem in question.
• Contextual: Cognitive computing should discover and extract
relevant information like location, time, and user details
pertaining to the problem based on sensory inputs such as
gestures, speech, and vision. Cognitive computing should ana-
lyze and process real-​time and near real-​time data.

DeepMind is a sophisticated computer that resembles the short-​term

memory properties of the human brain. This system has a neural net-
work connected to an external memory. It preservers information/​
queries as memories and then uses them later to understand and act
upon new data that it encounters on its own without any human inter-
vention or monitoring.
The Zeroth Cognitive Computing Platform is built upon the visual
and audio based cognitive computing capabilities that perfectly
resemble the actions and thoughts of a human brain. It enables the
system to accurately identify things, recognize and read handwritings,
recognize people all pertaining to specific contexts just like a human
brain. A few applications of the system could be the automatic adjust-
ment of the camera settings to accommodate to the natural light on
a bright sunny or cloudy day, and the adjustment of the microphone
setting to cancel out any disturbances or noise in the background to
help in improving the sound quality. It can also be applied to detect
the levels of stress and other feelings of a person from their voices
and send alerts.
 Smart Computing over IoT–Cloud 89

Watson is an advanced system with in-​built cognitive computing

capabilities such as dynamic learning, natural language processing,
and hypothesis generation and evaluation. It works on normal queries
that are asked by users in simple language and converts them into an
appropriate data language used for querying.
The information technology service giants such as TCS, Infosys,
Wipro, Cognizant, and HCL Technologies have been successful in
reducing human effort and cutting costs by developing their own tech-
nology platforms that incorporate automation, machine learning, and
artificial intelligence capabilities. Many tasks have been automated,
thereby generating more revenue without much of human effort
requirements. TCS’ automation platform is Ignio, HCL’s is Dry Ice,
Infosys operates Infosys Automation Platforms, Wipro manages
Holmes, and Cognizant has the Tizentto platform.
Cognitive computing is capable of minimizing the amount of traffic
from and to the Cloud in an IoT–​Cloud system by imparting intelli‑
gence to the edge devices. Devices can be equipped with capabilities
that can reduce energy consumption and improve performance and
privacy.

4 . 2 . 2 Und erlyin g Tec h n ologie s

As mentioned earlier cognitive computing is a confluence of many
underlying technologies such as natural language processing (NLP),
pattern recognition, data mining, sentiment analysis, machine
learning, neural networks, and deep learning. This section will pro-
vide a brief description of the technologies.
Natural language processing (NLP) [2] is a field of study that
helps in translating and interpreting human language by computers.
These computers basically work upon the natural human language
by analyzing and understanding the language, thereby being able to
extract meaningful information in a smart way. A piece of software
written by developers using the underlying NLP algorithms can help
understand the human language (speech and text) better and use it for
analysis. Some of the applications made possible due to NLP’s ability
to extract meaning from language based on the analysis of the hier-
archical structure of language are grammar correction, speech to text
convertor, and automatic language translators. NLP algorithms are
machine learning based algorithms that enable learning of rules of
execution by studying/​analyzing a predefined set of examples such as
90 Smart Computing over IoT–Cloud

books or a set of sentences, leading to a statistically generated infer-

ence. Here, the programmer is relieved from the burden of having to
write/​code the set of rules for analysis.
NLP has a set of open standard libraries that assist in real-​time
application development. Algorithmia is a model based on machine
learning that supports deployment and management of applications
without the need to spend efforts in setting up servers and infra-
structure. It helps a great deal in automating the machine learning
operations for an organization with simple API endpoints to the
algorithms, some of which are discussed below.
Apache OpenNLP is a toolkit employed to process text that is
written in the natural language and helps in development of services
that support proficient text processing actions. Common tasks that are
performed by NLP like language recognition, segmentation, parsing,
and chunking tokenization are supported by this open-​source library
that is based on machine learning.
Natural Language Toolkit (NLTK) is a collection of efficient libraries
for processing text in natural language (English). It is a platform that
supports symbolic and statistical NLP with programs written in Python.
NLTK has been found more suitable for teaching and research. It is also
suitable for empirical linguistics in Python, machine learning, artificial
intelligence, and retrieval of meaningful information.
Stanford NLP is a package of NLP software developed and
managed by the Stanford NLP group. The tools offered by the group
can be integrated into applications that require human language pro-
cessing, analysis, and interpretation requirements. Its use has been
extensive in the fields of academia and in industrial and governmental
organizations.
MALLET is a package that is open source and written in Java. It
supports statistical NLP with refined tools for various NLP activities
such as sequence tagging, numerical optimization, topic modeling,
information extraction, and beyond. One of the most easily under-
standable and relatable examples of NLP is the analysis of text for
the tone of message and then marking them as positive, negative,
or neutral as done in social media platforms such as Facebook or
Twitter. Tracking of topics that are currently trending and popularity
tracking of hashtags, auto filtering of offensive text and comments,
and chatbots are all efficient implementations of NLP.
Data mining is the process of excavating huge volumes of data
in order to draw inferences, patterns, knowledge, and informa-
tion that can be used to make improved business decisions, devise
 Smart Computing over IoT–Cloud 91

effective cost-​cutting strategies, and improve revenue. It involves

the application of certain mechanisms to help in finding anomalies
and correlations in larger data sets, thereby enabling detection of
outcomes [3]. It is one of the phases in the process of “knowledge
discovery in databases”. Data mining has six classes of tasks that are
performed during the process as listed below.

• Anomaly detection is a mechanism applied in order to examine

the data and detect any change or outlier or deviation that can
be used for further analysis and investigation.
• Association rule learning or also known as market basket ana-
lysis is a method of identifying relationships, patterns in the
data, and associated variables. For example, identification of
customer buying habits can help business in understanding fre-
quently bought items and items bought together. This can help
in developing efficient marketing strategies.
• Clustering is an action of identifying similarities in data, thus
leading to the detection of groups and structures in the data.
This grouping is done without any predefined/​known labels or
structures in the data.
• Classification is similar to that of clustering in that both the
methods perform grouping of data based on identified features.
However, classification is a supervised learning technique
wherein the data is categorized based on previously available
labels or structures, for example, classification of emails as
spam and valid.
• Regression is a predictive analysis mechanism wherein mod-
eling of data is done in order to assess the strength and relation-
ship between variables or data sets.
• Summarization is a representation of data sets in a compressed
and compacted way with the help of reports and visualizations
(graphs and charts).

Machine learning is a mechanism that imparts knowledge into

machines/​computer, thereby enabling them to act faster and effect-
ively. The ability to self-​learn and improve its response and functioning
based on the experience and with no external inputs (programming) is
the crux of machine learning built on the idea of artificial intelligence
[4]. There are basically three approaches toward machine learning
based on the type of stimuli/​signal and feedback that the learning
agent works upon.
92 Smart Computing over IoT–Cloud

Supervised learning is a mechanism wherein the learning agent

is provided some inputs, rules, and expected outcomes as examples
from which it is expected to learn and then apply them in the actual
operational environment.
Unsupervised learning is a mechanism wherein the learning agent
is not provided with any kind of prior knowledge and the agent is
given the liberty to act on the environment and generate outcomes on
its own based on its own analysis.
Reinforcement learning [5] involves a feedback/​ reward mech-
anism. The agent communicates with the real time dynamically chan-
ging environment in order to achieve a goal and at regular intervals
feedback/​rewards are presented based on the progress, which the
agent tries to maximize in its journey toward accomplishing the goal.
Neural networks are systems that take its root from a human brain.
They are similar to machine learning models in that they learn from
experience without the need for programming that is specific to the
task. However, neural networks are different in that they are able to
make intelligent decisions on its own, unlike machine learning where
decisions are made based on what it has learned only.
Neural networks have several layers that are involved in the refine-
ment of output at each level. The layers have nodes that are respon-
sible for carrying out the computations. It mimics the neuron in the
human brain wherein it collects and unites the input from the data and
assigns suitable weights which are responsible to either magnify or
diminish the input value pertaining to the task. An activation function
is assigned to a node that decides about the depth of progress that is
to be made by the signal through the various layers in order to affect
the outcome based upon the sum of the product of the input–​weight
pair. A neuron is activated in each layer when a signal reaches or
propagates through it.
Sentiment analysis finds its application in many well-​ known
platforms and applications. It is basically a text analysis mech-
anism that is used to detect the polarity (positive, negative, or neutral
opinion) in a given text, be it a long document, paragraph, or a few
sentences [6]. It basically helps in understanding human emotions
from the text. There are various types of sentiment analysis, and they
are described below.

• Fine-​grained sentiment analysis is used when the precision

of sentiment/​polarity is very important for the business. For
example, in a five-​star rating for a product or a review for a
 Smart Computing over IoT–Cloud 93

session, the polarity is recorded in five degrees (Excellent –​5,

good –​4, fair –​3, meets expectations –​2, below expectations –​1).
• Emotion detection is a method applied to identify human
emotions such as happy, sad, angry, or frustrated from a text
using a variety of machine learning algorithms.
• Aspect-​based sentiment analysis is a mechanism that helps
identify emotions expressed pertaining to specific features
or characteristics of a product or artifact. For example, when
customers review a phone, they may specify certain features of
the phone as being outstanding or abysmal.
• Multilingual sentiment analysis is a method applied on texts
written in a variety of languages. This involves a lot of effort in
terms of pre-​processing and resource requirements. Many tools
are also available online that can be used effectively, but for
more accurate and specific results, algorithms and techniques
must be developed indigenously.

Patten recognition is a mechanism applied to analyze and process

data in order to generate meaningful patterns or regularities. Pattern
recognition has its application in various areas such as image ana-
lysis, bioinformatics, computer graphics, signal processing, image
processing, and many more [7]. Pattern recognition techniques are of
three types as follows.
In statistical pattern recognition, patterns are described using
features and measurements. Each feature is represented as a point
in an n-​dimensional space. Statistical pattern recognition then picks
features and measurements that permit the pattern vectors to fit in
various groups or categories in the n-​dimensional space.
Syntactic pattern recognition uses basic subpatterns or simply
referred to as primitives that are utilized in making descriptions of the
patterns by organizing them into words and sentences.
Neural pattern recognition works on a system containing numerous
processors that are interconnected and enable parallel computing.
The systems are then trained on sample sets, enabling them to learn
from the given input–​output data sets, thereby enabling them to adapt
themselves to the changing data.
Deep learning enables machines to “learn by example.” A machine
acquires the ability to carry out classification by learning from
images, sounds, or text messages directly. Deep learning provides
unprecedented levels of accuracy that even exceeds normal human
ability. Data that are labeled are used to train the models using the
94 Smart Computing over IoT–Cloud

several layers in the neural networks. Deep learning will be discussed

in detail in Section 4.3.

4 . 2 . 3 Em p owerin g A n alytic s
In the introduction of this chapter, we had presented an apt defin-
ition for big data and cognitive computing. The huge volumes of the
big data pose huge problems too. Industries and businesses today are
overwhelmed with the amount of information that is accumulated for
analysis. However, the talent that is required to handle such data and
retrieve meaningful information for businesses is scarcely available.
The number of data scientists and analysts are not enough to keep
up with the ever-​increasing data volumes. Experienced specialists are
required in order to handle the available platforms and put them to
effective use.
A solution to this problem would be to increase the supply of
specialists by increasing the training programs offered to interested
people. On these grounds, an even more effective solution would be
to utilize the existing technologies and train machines/​computers
rather than human beings to manage the tools. This is made possible
with advancements in cognitive computing. The confluence of cogni-
tive computing, artificial intelligence, and machine learning can aid
both experienced and inexperienced staff to handle complex analytic
operations using the available tools and platforms. It also helps in
improving the accuracy and quality of the results. This accelerates the
process of analysis in real-​time and near real-​time data, thereby enab-
ling businesses to make real-​time decisions.
The capabilities that cognitive computing can empower big data
with are countless and very promising. The argument presented
above about the lack of sufficient talent with knowledge to handle
the big data platforms can be overcome with advancements in NLP.
With NLP in picture, employees who are not proficient in data/​infor-
mation processing and data languages that is required for analytics
activities can simply work on the platforms and tools with normal
interactions just as we do with other human beings. The platforms
can be equipped with the capability to transform normal language
into data queries and requests and respond with solutions or answers
in the same way as a natural language, enabling easy understanding.
This brings in much more flexibility into big data analytics.
Big data analytics empowered by cognitive computing has
accelerated the decision-​making process, accuracy, and productivity
 Smart Computing over IoT–Cloud 95

of many businesses with its tools and platforms. A few of them are
DeepMind, Watson, and Zeroth Cognitive Computing Platform,
which we have already discussed in Section 4.1.1. Following are a
few more tools that drive the analytics process in many businesses.
IBM Watson has partnered with an Indian health care organization’s
oncology department based in Manipal. It aids in making better
decisions in cancer treatment based on the previous knowledge that
it has acquired by going through a vast number of medical records,
patient history, and journals to arrive at an accurate diagnosis and
recommended treatments.
Wipro’s Holmes Platform mentioned in Section 4.1.1 is an artifi-
cial intelligence based platform. It has been adopted in the financial
sector in order to study, examine, and analyze a variety of structured
and unstructured data and support the decision-​making process of
an analyst. Tasks such as know your customer (KYC) and appraisal
pertaining to credit risk are automated with the help of Wipro Holmes.
The government of Andhra Pradesh has sought the predictive ser-
vices offered by Azure in order to identify students with a high prob-
ability of dropout from schools. The platform enables the collection
and analysis of data for specific information such as school infrastruc-
ture, teacher quality, training methodology, and also the social and
economic background. This enables the government in making better
decisions and allocate resources where in need.

4.3 DEEP LEARNING APPROACHES

Deep learning is a subset of machine learning in which the age-​old
traditional algorithms used to instruct the computers on the tasks
to be performed are equipped with capabilities to modify their own
instructions to improve the functionality of the computer. This
is the crux of artificial intelligence. While machines have great
advantages of accuracy and speed, they are limited by the program
parameters. In order to make the machines more competitive, they
should be made more adaptable like the human brain. Deep learning
is a mechanism that enables computers to process huge volumes of
data and learn from experience similar to humans. The algorithms
and mechanisms in deep learning perform tasks in a repetitive
manner, each time adjusting the parameters in order to achieve the
desired outcome. This section discusses a few approaches in deep
learning.
96 Smart Computing over IoT–Cloud

4 . 3 . 1 Ar ti fi cial Ne ural Networks ( A NN )

ANN is a deep learning approach that imparts artificial intelli-
gence into machines by mimicking the human brain and the nervous
system. Imagine that you have just hurt your index finger. The sen-
sory nerve in the hand immediately sends out signals (chain reaction)
that ultimately reaches your brain and tells that you are experiencing
pain. This is the basic idea behind the functionality of ANNs. The
ANN have a sequence of branching nodes which function in a similar
way to that of the neurons in the human body. Inputs are fed into the
input nodes, which then propagate the information into the series
of internal nodes, which process the information until the desired
output is generated.
Neural networks revolve around the two most vital components of
the brain, namely, the neuron and the synapse. The neuron is respon-
sible for processing the signals that it receives from other neurons.
The synapse connects the neurons enabling information to be passed
on from one neuron to another. In ANN, a node depicts the neuron
which receives the information and transforms the information by
performing a quantitative function, which is carried over to the next
neuron. In transit the connectivity lines (synapse) in turn apply its
own transformation function on the information and modifies it by
adding a constant value. This modification process happens by the
application of weights. The input from multiple synapses or connect-
ivity lines are collected, summed up, and then sent to the next node.
This node in turn adjusts the data by applying constants, thereby
modifying the data. This application of constants at the nodes is
called as node bias. Application of weights and bias to the input data
is important as this ascertains that the data is propagated properly
through the network. For a node to be able to propagate/​pass the data,
it must be activated. Activation of nodes happens when the output it
produces meets the threshold value that is set by the programmer,
after which the data will be passed on to the next node; otherwise
the node remains dormant. However, this single pass of informa-
tion to the final nodes in many cases might not lead to the desired
output. For example, the network might accidentally identify a cat as
a dog, which is not acceptable. To counter this, an algorithm called
the backpropagation algorithm is applied to the network, which uses
feedbacks to enable the adjustment of weights and biases and fine-​
tune the synapses until the result is agreeable or even almost correct.
The backpropagation algorithm is discussed in detail in Section 4.3.
 Smart Computing over IoT–Cloud 97

Figure 4.1 Artificial neural network.

Figure 4.1 shows a representation of the artificial neural network as

described above.

4 . 3 . 2 C o nvolution Neural Network ( CNN )

CNN is mostly applied to image processing problems and natural
language processing problems. Traditional neural networks make
no assumptions on the inputs and weights that are used to train the
models which are not suitable for images and natural language-​based
problems. CNN treats data as being spatial. As opposed to neurons
being connected to neurons in the preceding layer, they are instead
linked to neurons that are only close to it and all of them have the
same weights. This simplification enables to maintain the spatial
property of the data set. The simplification of an image is made pos-
sible via the CNN facilitating better processing and understanding of
the images [8].
The CNN architecture as shown in Figure 4.2 consists of multiple
layers that exist in a usual neural network. The CNN has additional
layers such as the convolution layer, pooling layer, ReLu (rectified
linear unit) layer, and a wholly connected layer. The ReLu layer
functions as an activation layer that ensures nonlinearity while the
data moves through each layer of the network. Absence of this layer
could cause the loss of dimensionality that is required to be maintained
with the data fed into each layer. The fully connected layer performs
classification on the data set. The convolution layer performs the most
important function in the network. It places a filter over an array of
98 Smart Computing over IoT–Cloud

Figure 4.2 Convolution network architecture.

Figure 4.3 Convolved feature map processing.

image pixels, leading to the formation of a “convolved feature map”

as shown in Figure 4.3.
This enables focus on specific features of the image, which might
be missed out otherwise. The pooling layer reduces the number
of samples of a feature map, causing a reduction in the number of
parameters to process, thereby enabling faster processing. This leads
to a pooled feature map. The pooled feature map is obtained by either
performing a max pooling (selects the maximum input of a par-
ticular convolved feature) or average pooling (calculates the average).
Ultimately, the model builds up an image of its own based on its own
mathematical rules. If unlabeled data are used, then unsupervised
learning methods can be used to train the CNN. Auto encoders enable
us to compress and place the data into a low-​dimensional space.
 Smart Computing over IoT–Cloud 99

Calculations are performed in the earlier layers of the CNN followed

by reconstruction with the other additional layers, leading to an up
sampling of all the data points at hand.

4 . 3 . 3 Recu rren t Ne ural Networks ( RNN )

RNN are used in many applications such as speech recognition, lan-
guage translation, and prediction of stocks. RNNs are used to model
sequential data. In order to understand this, consider a still snapshot
taken of a ball that is in motion over a period [9]. Now, from this pic-
ture we would want to predict the direction of motion that the ball is
moving in. Such a prediction with a single standstill picture of the ball
will be very difficult. Knowledge of where the ball had been before
the current picture is required for an accurate and real-​time predic-
tion. If we have recorded many snapshots of the ball’s position in
succession (sequential information over time), we will have enough
information to make better predictions. Similarly, audio is sequen-
tial information that can be broken up into smaller pieces of infor-
mation and fed into the RNN for processing. Textual data is another
type of sequential information that can be a sequence of alphabets or
words. RNNs work upon these kinds of sequential information to pro-
vide predictions based upon a concept called “sequential memory.”
Sequential memory is a mechanism that enables the human brain to
recognize patterns that occur in a sequence. This is implemented in
RNNs with the help of a looping mechanism that enables to pass on
earlier information in the forward direction for processing. This inter-
mediary information is represented as the hidden state, which depicts
previous inputs that affect the later states. An overview of a use case
can help better understand this looping mechanism in RNNs. Consider
building a chatbot that has a goal of classifying users’ intentions from
the text input provided. This can be handled by encoding the sequence
of texts using an RNN. The output received from the RNN is then
fed into a standard feed-​forward neural network, which will then per-
form the classification of the intentions. So, consider the following
example scenario where user inputs “What is the time?” Figure 4.4
represents the steps 1–​5 as mentioned in the algorithm.
INPUT: WHAT TIME IS IT?

1. Break the sentence into individual words as WHAT TIME

IS IT?
100 Smart Computing over IoT–Cloud

Figure 4.4 Working of a recurrent neural network.

2. Feed the words into the RNN in a sequential manner. Start by

feeding in the first word “WHAT” into the RNN. The RNN
encodes “WHAT” and produces and output.
3. Next we feed the word “TIME” with a hidden state from the
previous step. The hidden state represents information from
all the previous steps. Now, the RNN has information on the
words “WHAT” and “TIME.”
4. Repeat the process for all words until the final step that feeds
“?” with all previous state information and produces the final
output.
5. Now, since the final stage’s (stage 5) output consists of infor-
mation from all other stages, the output of stage 5 can be
taken and provided as input to the feed-​forward network that
performs a classification of intent on the input.

The RNN, however, suffers from a problem, which is short-​term

memory and vanishing gradient, which is the side effect of the back-​
propagation methodology in RNNs. In simple terms, the final output
that occurs after sequentially passing in all the input chunks barely
carries the memory/​information of the output from stage 1, thereby
affecting the final output (e.g., the words “WHAT” and “TIME” may
not even be considered in predicting the users’ intent (short-​term
memory loss)). Output based on merely “IS” and “IT” may be very
ambiguous and unclear. This diminishing information can be handled
with advanced RNNs, namely long short-​term memory (LSTM) and
gated recurrent unit (GRU).
RNN differs from CNN in that, CNN is a feed-​forward network
used to filter spatial data while the recurrent neural network (RNN)
feeds data back into itself, thereby being the best candidates for
sequential data. A CNN can recognize patterns across space whereas
RNN can recognize patterns over time.
 Smart Computing over IoT–Cloud 101

4.4 ALGORITHMS, METHODS, AND TECHNIQUES

Section 4.3 has provided us a brief understanding about deep learning
and its approaches. This section will discuss about a few deep learning
methods and AI algorithms that are used widely. Following are a few
deep learning methods [10] that are commonly used in applications.
Back propagation is a method of training the neural networks that
works on the principle of the supervised learning methodology. As
explained in Section 4.2.2, neural networks have several layers that
process the input to achieve the desired output. While training the
model, we provide arbitrary weights at the input that depicts the rele-
vance of each node. The desired output is also specified; however,
as these weights are randomly assigned, the weights add up to some
amount of error during the process and causes a great difference at
the output layer. Backpropagation helps in fine-​tuning the weights
by reassigning it to an approximate value based on the difference
inferred between the actual and desired output. The iterations are
repeated until a suitable weight is achieved for the model with min-
imal error value. The pseudo code is provided below.

1. Values (x,y) are supplied to the input layer nodes.

2. Inputs are assigned arbitrary weights (w1, w2).
3. The values propagate forward through the model passing all
the way through the hidden layer nodes to the output nodes
where the output is calculated for every hidden node and the
output nodes.
4. The error value (difference between the actual output and the
required output) is calculated.
5. Backpropagation from the output nodes to the input nodes
is done and, at each node, the weights are adjusted and
reassigned so that the error value in step 4 is reduced.
6. The process is repeated for several iterations until the actual
output and desired output have minimal or no error (difference
in value).

Figure 4.5 shows the working of a back-​propagation network. The

numbers in the figure represent the steps specified in the pseudo code.
The process of creating a simple neural network and training it
using the back propagation algorithm is presented in reference [11].
Stochastic gradient descent is a method that helps in sampling
huge volumes of data by randomly selecting data points and sampling
them, thereby reducing the amount of computation required. It is
102 Smart Computing over IoT–Cloud

Figure 4.5 Working of a backpropagation network.

Figure 4.6 Stochastic gradient descent.

based on the unsupervised learning mechanism. A simplified explan-

ation to this method can be given with reference to Figure 4.6. The
objective is to find a point x such that the objective function y amounts
to zero or minimum. The pseudo code is given below. In the parabola,
the lowest point is obtained at x=1. However, the objective is to find a
value x such that y is minimum (y= –​3 where x=0 or 2).

1. Compute the slope for y (objective function also referred to as

the gradient descent function at iterations) with respect to the
parameters/​features.
2. Assign initial value for the parameters at random (differentiate
y with respect to x).
3. Revise the gradient descent function with these new parameter
values.
 Smart Computing over IoT–Cloud 103

4. Calculate the step size for every feature as step size = gradient
* learning rate. Learning rate should be kept to a minimum so
that on its journey down the slope we do not want to miss out
the minimum point by taking larger strides.
5. Determine new parameter values as new parameter value = old
value –​step size.
6. Continue the process by repeating steps 3–​5 until the gradient
reaches a value 0 or close.

Transfer learning [12] is a method adopted to train models in layers

by adopting the convolution mechanism. The last few layers tend to
be more specific to the data fed as input while the starting layers are
more generic pertaining to simple patterns. For example, in a model
provided with a training dataset, the early layers might be looking for
eyes, ears, and mouth while the later layers may be looking for dogs,
humans, and cars.
Following are a few AI algorithms that are most used in applications.
Logistic regression is a classification algorithm working on the
principle of supervised learning. It forecasts the probability of a
dependent or target variable. The dependent variable has only two
outcomes coded as 1 (success/​yes) and 0 (failure/​no). It is the sim-
plest mechanism used in classification problems such as illness pre-
diction (cancer and diabetes) and spam email identification.
Naive Bayes algorithm is another classification method based on
the supervised learning mechanism. The central idea of the naïve
Bayes classifier is the Bayes theorem. The classification has two
phases, namely the learning phase (model trained on a given dataset)
and evaluation phase (performance testing). Given the example of
weather conditions and playing a sport, the goal is to determine the
probability of playing the sport and also determine if the players will
play the sport or not (yes or no) with respect to the weather. The Bayes
algorithm is given as follows:
where

• P(H) is the prior probability of H (yes/​no), that is, probability of

hypothesis being true.
• P(D) is prior probability of D (for example, D can be the wea-
ther conditions such as sunny or rainy).
• P(H|D) is probability of H given D and is called the posterior
probability.
• P(D|H) is probability of D given H and is also called the pos-
terior probability.
104 Smart Computing over IoT–Cloud

Problem: The players play if the weather is rainy. Determine if the

statement is true or false. Refer Table 4.1 for values. Table 4.2 shows
the frequencies of the classes and labels. Naïve Bayes classifier
determines the probability of an event as follows.

1. Determine the prior probability for the known class labels

(for example, yes/​no) and for each attribute (D, i.e., (Rainy,
Sunny)) as shown in Table 4.3.
P(yes) = 2/​5 = 0.4, P(no) = 3/​5 = 0.6, P(Sunny) = 3/​5 = 0.6,
P(Rainy) = 2/​5 = 0.4.

2. Find the posterior probability for each attribute (D, i.e.,

(Rainy, Sunny)) and for every class label (yes,no) as shown in
Table 4.4.
P(Sunny|No)=2/​3=0.67, P(Rainy|No)=1/​3=0.33,
P(Sunny|yes)=1/​2=0.5, P(Rainy|Yes)=1/​2=0.5

3. Substitute the values in Bayes formula and compute posterior

probabilities. Probability that players won’t play when the
weather is rainy is given by
P(No|Rainy) = [P(Rainy|No)*P(No)]/​P(Rainy) = (0.33*0.6)/​
0.4=0.495
and the probability that the players will play given that the
weather is rainy is given by
P(Yes|Rainy) = [P(Rainy|Yes)*P(Yes)]/​P(Rainy) = (0.5*0.4)/​
0.4=0.5
4. Evaluate the class with higher probability, such that the input
belongs to the class with higher probability. In the example,
the probability that the players will play when the weather is
rainy has a slightly higher probability in comparison to that of
the probability the players will not play if it rains. Hence, we
can conclude that the hypothesis/​problem statement is true.

A similar example [13] as provided above can be referred as a

guidance to the implementation of the naïve Bayes classifier. The
author has used a simple data set with three columns, namely, wea-
ther, temperature, and play for effective demonstration.
Support vector machines (SVM) is a machine learning algorithm
that follows the supervised learning mechanism. It is widely used in
classification problems where the goal is to find a hyperplane that best
segregates all the data points into two different categories as shown in
 Smart Computing over IoT–Cloud 105

TABLE 4.1
Naïve Bayes Data Set
Weather Class Label
Sunny No
Sunny No
Rainy Yes
Rainy No
Sunny Yes

TABLE 4.2
Naïve Bayes Frequency Table
Weather No Yes
Sunny 2 1
Rainy 1 1

TABLE 4.3
Naïve Bayes Prior Probability
Weather Yes No
Sunny 2 1 =3/​5=0
Rainy 1 1 =2/​5=0.4
Total 3/​5=0.6 2/​5=0.4

TABLE 4.4
Naïve Bayes Posterior Probability
Weather Yes No Posterior Probability Posterior Probability
(Yes) (No)
Sunny 2 1 2/​3=0.67 ½=0.5
Rainy 1 1 1/​3=0.35 ½=0.5
Total 3 2
106 Smart Computing over IoT–Cloud

Figure 4.7 Support vector machine.

Figure 4.7. Support vectors are nothing but data points themselves that
are closer to the hyperplane whose removal can cause a great change or
shift in position of the hyperplane. The success of SVMs is determined
from the distance that exists between the hyperplane and the nearest
data point, which is called as the margin (greater distance means
effective classification). When a clear hyperplane cannot be identified,
a 3D view of the data points can help in obtaining the hyperplane,
which is done through a mechanism called kernelling. SVM finds its
application in many areas such as in the cancer and neurological dis-
ease diagnosis and many other research related to health care.
For further understanding, readers can refer to the example in refer-
ence [14], which shows the implementation of support vector machine
in Python using scikit-​learn. The data set used for the implementation
is the caner data set, which can be downloaded from the UCI Machine
Learning Library. The data set comprising 30 features classifies cancer
into two classes (target), namely malignant and benign.
Following are a few machine learning techniques that are used
commonly in cognitive computing [15].
Random Forest as the name suggests is a collection of decision
trees that when presented with an entity for classification performs an
analysis and presents a class to which the entity might belong. Each
tree votes for a class independently and a unanimous decision is made
 Smart Computing over IoT–Cloud 107

Figure 4.8 CART.

by choosing the class that has the largest number of votes. The deci-
sion trees are implemented as a forest in the following steps.

1. If the training data set holds N cases, then a sample space

containing N cases are selected randomly with replacement,
which will be used to grow the decision tree.
2. For an input with M variables, m variables lesser than M are
selected randomly and the best possible split value is used to
split the node. This m value remains constant throughout the
process of growing the forest.
3. Every decision tree is grown to the maximum size possible
without pruning.

CART stands for classification and regression tree. The non-​terminal

nodes (root node and intermediate nodes) signify a single input vari-
able (x) and a splitting point on the variable. The terminal nodes (leaf
nodes) signify the output variable (y). The model predicts the output
by walking through the decision tree paths that satisfy the conditions
and arrives at the specific output. For example, in Figure 4.8, to know
if an unmarried person aged above 30 will buy a minivan or a sports car
will follow the path: Over 30 yrs (yes) → Married (No) → Sports car.

4.5 CASE STUDIES

4 . 5 . 1 Hea lth C are
As the average life expectancy of humans increases, health care
is evolving into multiple specialties and unlimited facets. In fact,
108 Smart Computing over IoT–Cloud

health care is a multi-​trillion-​dollar industry with technology that has

changed the way medical professionals work by enabling accurate
diagnosis and improved care. It also saves lives and lessens suffering.
In health care, deep learning is implemented in every stage of the
workflow to elevate the standard of care and quality of life. One of the
most important challenges in health care is to get the right treatment
to patients as soon as possible and as efficiently as possible. In inten-
sive care units (ICUs), unrelenting attention is necessary with large,
incessant data to be processed in order to take critical decisions as fast
as possible. Researchers in MIT’s Computer Science and Artificial
Intelligence lab have developed a deep learning tool called ICU
Intervene, which will predict whether the patients will require cer-
tain ICU treatments. This model utilizes hourly measurements of very
important signs like blood pressure, heart rate, and glucose levels
along with patient data like age and gender in order to predict neces-
sary treatments.
Another deep learning model by a startup Corti analyzes phone
calls from emergency services. The audio recorded is analyzed and
helps emergency personnel to identify if a caller is about to undergo
cardiac arrests within a minute. This is also a time-​sensitive decision-​
making process. LexiconAI is a speech recognizing mobile app that
can capture medical information with doctor–​patient interactions.
This enables automatic filling of electronic medical records, thereby
enabling doctors to interact more with patients. Millions of med-
ical scans and tissue biopsies are going digital every year. Proscia, a
startup uses deep learning with 99 percent accuracy to classify skin
pathologies. SigTuple, a startup, has developed an AI microscope that
can analyze blood and other fluids of the body. It can analyze the
digital images either in the Cloud or at the microscope itself, when
compared to digital scanners. SigTuple does it in a fraction of a cost.
To help people with impaired vision, Aira a startup uses AI-​guided
service used on a smart glass, which helps people navigate and per-
form their daily routines easily. Aira connects with multiple service
agents for assistance. Agents can see what the blind person sees,
enabling assistance in just about anything. Aira has a forward-​facing
camera along with the ability to connect to a smartphone to tap into the
GPS and connectivity. To help customers with more complex tasks,
human agents are accessible at the tap of a button on the glasses or
the app. The agents communicate directly with Aira users via phone,
providing real-​time visual information with audio support. Complex
image recognition and natural language processing tasks are sent to
 Smart Computing over IoT–Cloud 109

the Cloud for inference. Georgia Tech has introduced an AI prosthetic

hand that assisted jazz musician Jason Barnes play piano for the first
time in five years. The prosthesis uses electromyogram sensors to rec-
ognize muscle movement and allows for individual finger control.

4 . 5 . 2 S m a r t Home
Like in health care, cognitive computing plays a key role in smart
homes. Applying machine learning enables smart homes to scale
effectively and efficiently. Predictive maintenance is the first key
benefit of Cognitive Smart Homes. It can provide critical informa-
tion to the building owners about equipment failure well ahead in
time. Nest Labs’ thermostat can also help lessen energy expenses
by using sensors that identify the number of occupants in a room by
automatically adjusting the temperature. Philips smart light bulbs can
turn on and off by detecting occupants in the room. Smart building is
connected to smart grid/​city and they can listen or talk to each other,
thereby making it self-​aware. Voice-​activated systems contain vir-
tual assistants that can learn and personalize the smart home to their
occupants. Smart TVs access content through internet along with
voice and gesture support. Yale Smart locks and garage door openers
can grant or deny access to visitors. It can automatically unlock the
door for its occupants. Smart security cameras enable residents to
observe their homes while they are away. Smart sensors can recognize
the difference between residents, pets, visitors, and burglars and alert
the authorities. Automation of pet care can be done with connected
feeders. Plants at home and gardens can be watered by attached
timers. Monitors of household systems can detect an electric surge,
failures in water system, or freezing pipes and immediately turn off
appliances so that the homes are safe. Smart refrigerators monitor the
expiry date, make a shopping list, or suggest a recipe based on the
contents of the fridge. Smart coffee makers can prepare a fresh cup of
coffee. The next logical step of improving the accuracy and capability
of IoT sensor driven systems is infusing human awareness via cogni-
tive computing and infuse intelligence into these devices.

4 . 5 . 3 Ma nuf ac turin g
Industrial firms are constantly seeking to bring down operational
costs and machine downtime. Cognitive computing offers predictive
110 Smart Computing over IoT–Cloud

maintenance solutions that can accurately predict equipment failure.

By reducing machine downtime, cognitive computing helps indus-
trial companies to work smarter and safer at the same time, redu-
cing operational expenditure. Predictive maintenance solution from
Baker Hughes GE is a deep learning-​based model designed to predict
equipment failure as early as two months in advance.
In early twentieth century, cars were sold with essentially no
options. More than a century later, almost 99 percent of vehicles
ordered are unique with customizations. In other words, every order
is a custom, so the logistics team in the automobile industry must
take in millions of raw parts a day from thousands of suppliers and
different locations and organize them into custom-​part trays for every
customer order. Multiple models of cars can be assembled at the same
time and a car is output every minute and the production line can
never stop. Achieving this high level of accuracy in factory logistics
requires AI-​powered industrial logistics. Smart robots in BMW are
autonomous in nature, which pick up the car trays and move them
to the manufacturing areas. These smart robots perform perception,
3D human pose estimation, localization, actuation, and path planning
performing trillions of operations per second of performance.

4 . 5 . 4 Reta i l Sec tor

Amazon Go is a prime example of the confluence of Cloud and IoT
put into action in the retail sector. It is the consumer store of the
future where customers can simply walk into the store, take what they
want, and come out. After a few minutes, users get a bill for all things
they have taken from the store. Amazon Go took three years just to
reach the proof-​of-​concept stage and another three years of rigorous
testing and development to reach production stage and opening up
the product to customers. It belongs to a class of NP-​complete algo-
rithmic problem.
The heart and soul of the Amazon Go Store is computer vision-​
based machine learning. It is a highly scalable system that can effect-
ively track and identify the objective of all the customers in the store.
Amazon Go had to solve six fundamental problems to provide this
seamless experience. It is called as the Just Walk Out Technology.
They are as follows.
Sensor Fusion represents the problems involved in combining
signals coming from different IoT devices in real time. This is very
 Smart Computing over IoT–Cloud 111

important because if two or more people come close together a

problem of tangled state occurs. It means that the AI system has low
confidence of who took what and who is who. Another problem is
occlusion where people or objects could be hidden from the view.
Calibration represents the problem associated with each IoT device
knowing its location precisely in the store. There could be items that
are very similar, say, the different flavors of tea from the same brand.
Lighting conditions in the store and deformation can change the looks
of the item.
Person Detection involves the ability to track and identify indi-
vidual persons throughout his or her stay in the store. Here it is highly
important that the labels are preserved across the video when the cus-
tomer is moving across the store.
Object Detection involves the ability to identify all the items that
are being sold.
Pose Estimation represents the ability to identify each customer’s
activity near the shelf, especially monitoring the movements of
their hands.
Activity Analysis involves checking if the customer has returned/​
replaced the item on the shelf or picked it up.
For the computer vision to be successful, it needs Cloud for com-
putation and the magic will not happen until the video feed does not
reach the Cloud quickly. Here the IoT devices play a vital role. It
does prepossess before streaming it to the Cloud, thereby reducing
the video stream size using novel methods such as compression using
video codecs. The IoT devices must be resilient enough to have a
fail-​safe mechanism in case of device failure or network failure such
as a backup mechanism to kick in to ensure that the stream reaches
the Cloud.

REFERENCES
1. Danish Contractor and Aaditya Telang, Applications of Cognitive
Computing Systems and IBM Watson, Springer 2017.
2. M. Filgueiras L. Damas N. Moreira, and A. P. Tomás, “Natural
Language Processing,” EAIA ‘90, 2nd Advanced School in
Artificial Intelligence Guarda, Portugal, October 8–​12, 1990.
3. Chun-​Wei Tsai, Chin-​Feng Lai, Ming-​Chao Chiang, and Laurence
T. Yang, “Data Mining for Internet of Things: A Survey,” IEEE
Communications Surveys & Tutorials, 16, 77–​97, 2014.
112 Smart Computing over IoT–Cloud

4. Mario Bkassiny, Yang Li, and Sudharman K. Jayaweera, “A

Survey on Machine Learning Techniques in Cognitive Radios,”
IEEE Communications Surveys & Tutorials, 15, 1136–​1159, 2013.
5. Akira Notsu, Katsuhiro Honda, Hidetomo Ichihashi, and Yuki
Komori, “Simple Reinforcement Learning for Small-​ Memory
Agent,” 10th International Conference on Machine Learning and
Applications and Workshops, 2011.
6. Neha Raghuvanshi and J. M. Patil, “A Brief Review on Sentiment
Analysis,” 2016 International Conference on Electrical,
Electronics, and Optimization Techniques (ICEEOT), 2016.
7. M. Basu, H. Bunke, and A. Del Bimbo, “Guest Editors’
Introduction to the Special Section on Syntactic and Structural
Pattern Recognition,” IEEE Transactions on Pattern Analysis and
Machine Intelligence, 2005.
8. Guangxin Lou and Hongzhen Shi, “Face Image Recognition
Based on Convolutional Neural Network,” IEEE 2020.
9. Mohammad Reza Mohammadi, Sayed Alireza Sadrossadat, Mir
Gholamreza Mortazavi, and Behzad Nouri, “A Brief Review
over Neural Network Modeling Techniques,” IEEE International
Conference on Power, Control, Signals and Instrumentation
Engineering (ICPCSI), 2017.
10. S. Durga, Rishabh Nag, and Esther Daniel, “Survey on Machine
Learning and Deep Learning Algorithms used in Internet of Things
(IoT) Healthcare,” 3rd International Conference on Computing
Methodologies and Communication (ICCMC), 2019.
11. “Backpropagation in Neural Networks: Process, Example & Code.”
Neural Network Concepts. https://​machinelearningmastery.com/​
implement-​backpropagation-​algorithm-​scratch-​python/​.
12. Ling Shao, Fan Zhu, and Xuelong Li, “Transfer Learning for
Visual Categorization: A Survey,” IEEE Transactions on Neural
Networks and Learning Systems, 26, 1019–​1034, 2019.
13. www.datacamp.com/ ​ c ommunity/ ​ t utorials/ ​ n aive- ​ b ayes-
scikit-​learn
14. www.datacamp.com/ ​c ommunity/ ​t utorials/​s vm-​ c lassification-​
scikit-​learn-​pythony
15. S. Athmaja, M. Hanumanthappa, and Vasantha Kavitha, “A
Survey of Machine Learning Algorithms for Big Data Analytics,”
International Conference on Innovations in Information,
Embedded and Communication Systems (ICIIECS), 2017.
 5

CO N CLUS I O N

5.1 SUMMARY
This section concludes the book and proposes the future outlook,
which can be explored further to initiate new research avenues. This
book intended to cover requirements, challenges, technologies, and
methods to be adapted in order to bring about the convergence of
IoT and Cloud for smart computing. The main topics covered in this
book are overview of IoT in terms of architecture and use cases, IoT
application development and protocol requirements, opportunities
and techniques for IoT–​Cloud convergence, and smart computing
requirements through learning techniques. The chapter-​wise details
are summarized below.
The advancements in tiny smart devices, IoT, and Cloud com-
puting have laid the basis for a converged service network called IoT.
The seamless communication and contextual services are provided by
IoT using smart objects. An overview of IoT covering the technical
building blocks and the usage of IoT in business have been presented
in Chapter 1. It also presents the IoT layered architecture. The chapter
proceeds further by providing an overview of smart computing and
discusses a few use cases. The chapter concludes with a discussion on
the IoT design issues and challenges.
Due to rapid advancements in e-​commerce and m-​Commerce as
well as availability of Internet at faster and cheaper rate, the use of
mobile applications and IoT-​enabled applications is increasing at
faster rate. IoT solutions have become an integral part of everyday
life and IoT application development is becoming a potential business
proposition. Smartphones, sensors, cameras, and RFID are the major
components in all IoT use cases. Minimum human intervention for
the exchange of information and interconnecting devices using wired
and wireless communication medium are the key steps in building
IoT applications. Chapter 2 begins with a brief description about the

113
114 Conclusion

main principles to be taken into consideration before developing IoT

applications. It then explains the IoT application development phases
using a novel C model. Later, the chapter introduces the reader to the
various wireless technologies that are suitable for IoT application such
as Zigbee, WiFi, and Bluetooth. The chapter also provides a detailed
explanation about the protocol stack that is used for IoT application
development, which are provided by various working groups. Finally
the chapter concludes with a brief explanation of different electronic
platforms and their comparison.
In small standalone consumer applications like the smart home and
care for disabled and elders, the capabilities required such as pro-
cessing power, storage space, and computing can be facilitated by the
participating devices (mobile phones, Raspberry Pi kits) themselves.
The resource requirements for such applications are lower. However,
for larger commercial real-​time applications such as health care, and
transportation, the demand for resources is huge due to the enormous
data that is collected and also the complexity of computations that are
involved in the system. The demand for resources (storage, servers,
and processors) in such huge applications many a times cannot be
satisfied by the participating devices alone. This is where Cloud
comes in handy by catering to larger business needs. In Chapter 3,
we primarily discussed about how Cloud plays a vital role in real-
izing IoT to its full potential. The opportunities and challenges that
are encountered when the IoT–​Cloud convergence comes into pic-
ture are also discussed in this chapter. It provided a bird’s eye view
of the architecture for enabling this convergence. It proceeds further
by elaborating on data offloading and computing from the IoT per-
spective. The concept of dynamic resource provisioning for IoT with
Cloud is also explained. Finally, the chapter introduced the various
test beds and technologies with respect to IoT–​Cloud convergence
application development.
Smart computing in the context of IoT–​Cloud leads to the introduc-
tion of advanced computing capabilities including real-​time awareness
of the surrounding environment and analytics (of the big data detected
from various sensors and devices) into the IoT applications by lever-
aging the capabilities of Cloud. This enables incorporating intelli-
gence into the decision-​ making process with advanced resources
made available through the Cloud. Chapter 4 provided an insight into
big data analytics and cognitive computing capabilities and the under-
lying technologies. The chapter further explained the various deep
learning approaches that help realize smart computing in a IoT–​Cloud
 Conclusion 115

system. A few major case studies under various domains where smart
computing over IoT–​Cloud is used have been elaborated in the chapter
enabling the readers acquire a clear and vivid picture of the usage of
smart computing over IoT–​Cloud.
The revolution in different technologies like Cloud computing,
IoT, cognitive computing, big data analytics and deep learning has
led to the emergence of smart application and efficient services with
less human intervention. This convergence of different technologies,
especially Cloud computing, has accelerated and eventually opened
up many personal and business opportunities. This chapter provides
a summary of each chapter. The chapter also discusses the issues and
challenges involved in developing an IoT–​Cloud application. It fur-
ther provides future outlook of Cloud and IoT in several domains,
thereby laying a foundation for further research.

5.2 ISSUES AND CHALLENGES

As stated earlier, the convergence of IoT and Cloud opens up numerous
opportunities, It also raises new issues and challenges, which arise
during the development of IoT applications or connecting these
applications to Cloud. IoT devices suffer from a scarcity of resources,
which includes limited memory, battery power, and computational
power. These devices are equipped with wireless transmitters to ini-
tiate communication and a primitive embedded operating system is
installed on these devices. The following is a summary of issues and
challenges [1,2]:

(a) Low-​speed networks: For connecting IoT devices to the

gateway, there is low-​speed network connectivity for data
transfer. However, data transfer is huge and there is a need of
lightweight data transfer algorithms for efficient data transfer.
(b) Scalable data handling algorithms: Data generated by het-
erogeneous IoT devices are finally aggregated at a central
data storage platform and scalable data handling algorithms
are required at the server side. The required operations to be
performed on the data are sorting, analyzing, and performing
machine learning related tasks like prediction or recommen-
dation. These algorithms at the server side need to be scalable
in order to support huge numbers of devices connected to IoT
applications, which are located at distributed geographical
locations.
116 Conclusion

(c) Information fusion: In emerging IoT applications like health

care and agriculture, the data comes from multiple types of
sensors and there is a need of efficient information fusion
techniques for multi-​sensory data. In applications like remote
sensing and geographic information system (GIS), the rele-
vance of multi-​sensor data poses a big challenge and therefore
reliable information fusion techniques are needed.
(d) Security: Due to distributed and decentralized nature of IoT
applications, they are more prone to attacks at various layers.
Possible places of attacks are at devices, communication link,
and at the Cloud storage, which include attacks on integrity,
confidentiality, privacy, and identity. To resist these various
types of attacks, we need lightweight security algorithm not
for detecting the attacks but for their prevention. Security by
design approach is recommended to address this challenge.
(e) Availability or downtime: Critical IoT-​enabled applications
like health care or defense do not afford downtime. A single
point of failure situation needs to be avoided during the con-
vergence of IoT and Cloud. Dynamic load balancing solution
should be in place for adaptive load transfer from failure node
to active node in order to avoid downtime.
(f) Quality of service: Quality of service parameters for IoT
applications includes latency, packet delivery ratio, and
energy consumption and their requirement varies from one
use case to another. The quality of service requirement
and feasibility analysis should be carried out in the design
part of application development to meet to the customers’
requirements.
(g) Protocol development: Protocol development in IoT suffers
from lack of standardization. So the selection of appro-
priate protocol is very important from the perspective of both
developers and customers. There should be also directed
efforts in order to standardize IoT protocols so that designers
and developers will have freedom to select from a pool of
protocols.

5.3 FUTURE OUTLOOK

The focus of this book is to put forth advances in the field of IoT
and Cloud convergence. There are still many areas that need further
 Conclusion 117

attention and have prospective applications in future. Convergence of

IoT and mobile Cloud for health care applications like remote telem-
etry is an important area which can be researched further.
Next generation mobile and wireless technologies will have greater
impact on the technologies and protocols used for the convergence of
IoT and Cloud [3]. Utmost care has to be taken to ensure interoper-
ability and compatibility in both upward and backward directions.
Due to scalable and efficient features of Cloud, it has become a
potential candidate in the application of IoT devices and storage and
processing of big data generated from IoT devices. However, data
privacy is at risk as these IoT devices are connected through several
communication technologies that are both wired and wireless. In add-
ition to this, the corresponding operating system on the smartphone
is also prone to data privacy leaks. Therefore, there is a need for new
frameworks for privacy-​aware data aggregation to ensure security and
privacy of next generation smartphone technologies used in integra-
tion of IoT to Cloud.
When IoT–​Cloud convergence is used for use cases like remote
patient monitoring, remote telemetry, or monitoring old age people
remotely, then intelligent computing at Cloud side requires more
advanced techniques, which include cognitive computing for data
science and quantum computing for personalization, which need fur-
ther research. Furthermore, the potential of quantum computing needs
to be extended for big data analytics for providing secure platform
for communication and cyber security and enhancing computation
power. Weather forecasting is another interesting use case, which
can enhance food quality, production, and retail sales. Quantum
computing can contribute to this field to a great extent by managing
complex arithmetic at a faster rate and in building climate models
to predict global warming and greenhouse gases. This can help the
society in preventing global environmental disasters by initiating pre-
ventive measures. Further, there is also a need of generic framework
that will integrate or converge multiple combinations like IoT and
Cloud, IoT and Big Data on Cloud, or any other suitable combination
of these three technologies.
In addition to this, there are other allied areas like experimenting
and validating data science techniques on real-​time data posted on the
Cloud, mapping various machine learning techniques with respect to
the nature and type of data set, and proposing appropriate tools for
Cloud data analytics.
118 Conclusion

REFERENCES
1. P. N. Mahalle and P. N. Railkar. Identity Management for Internet of
Things. River Publishers, Wharton, TX, 2015.
2. P. Mahalle, “Identity Management Framework for Internet of
Things.” PhD Dissertation. Aalborg University, Denmark, 2013.
3. M. S. Hossain, C. Xu, Y. Li, A.-​S. K. Pathan, J. Bilbao, W. Zeng,
and A. El Saddik, “Impact of Next-​Generation Mobile Technologies
on IoT-​Cloud Convergence,” Communications Magazine, 55(1),
18–​19, 2017.
 I N D EX

6LowPAN 28 Azure IoT Central 53

Azure IoT Edge 53
Advanced message queuing Azure IoT hub 53
protocol (AMQP) 32 Azure IoT Solution Accelerators 53
Aira 108 Azure maps 54
Algorithimia 90 Azure RTOS 54
Amazon Go 110 Azure Sphere 53
Amazon Web Services (AWS) 51 Azure Time Series Insight 53
Anomaly Detection 91
Apache Open NLP 90 Back Propagation 101
Application level offloading 66 Baker Huges GE 110
Application Programming Interfaces BeagleBone 37
(APIs) 9 Big Data Analytics 85
Arduino 35 Bluetooth 26
Artificial Neural Networks Bluetooth Low Energy 26
(ANN) 96 Body Sensor Network 8
Association Rule 91 BUTLER 13
Authentication 77
Authorization 77 CART 107
AWS IoT Analytics 52 CASAGRAS 10
AWS IoT Core 52 Classification 91
AWS IoT Device Defender 52 Cloud Bursting 68
AWS IoT Device Management 52 Cloud Computing 42
AWS IoT Events 52 Cloud of Things 63
AWS IoT Greengrass 52 Cloudlets 62
AWS IoT SiteWise 52 Clustering 91
AWS IoT Things graph 52 Cognitive Computing 85
Azure Digital Twins 53 Com4Innov 80
120 Index

Constrained application protocol International Telecommunication

(CoAP) 29 Union 1
Context 25 Internet of Things (IoT) 1
Controller Area Network (CAN) 35 Intrusion Detection 77
Convergence 25 IoT-​ A 8
Convolution Neural Networks
(CNN) 97 KETI 80
Corti 108
LexiconAI 108
Data Mining 90 Logistic regression 103
Datagram Transport Layer Security Long short-​term memory (LSTM)
(DTLS) 30 100
Deep learning 93 LoRaWAN 16
DeepMind 88
Deployment Models 43 Machine learning 91
Device-​centric edge-​based 76 MAGNET 10
Dynamic resource provisioning 68 MAGNET Beyond 10
MALLET 90
End-​to-​end 76 Message Queuing Telemetry
European Union Network and Transport (MQTT) 31
Information Security Agency Method level offloading 66
(ENISA) 72 Micro Data Centers 62
Extensible messaging and presence Mobility 19
protocol (XMPP) 33 Multi-​Access Computing or Mobile
Edge Computing (MEC) 61
Feature map 98
FIESTA-​IoT 80 Naïve Bayes 103
Firewall 76 Natural Language processing (NLP)
Fog Computing 61 89
FreeRTOS 51 Natural Language Toolkit (NLTK) 90
Neural Networks 92
Gated recurrent unit (GRU) 100 Node bias 96
Google BigQuery 54
Google Cloud DataFlow 54 Pattern Recognition 93
Google Cloud ioT Core 54 Privacy-​preserving 78
Google Cloud MLEngine 54
Google Cloud Pub/​Sub 54 Radio Frequency Identification 2
Random Forest 106
Heterogeneity 19 Raspberry Pi 36
High Definition Multimedia Recurrent neural network (RNN) 99
Interface (HDMI) 36 Regression 91
Reinforcement learning 92
IBM Watson 89 Representational State Transfer
iKAAS 56 (REST) 30
 Index 121

Scalability 19 Transfer learning 103

Sentiment Analysis 92
Service Models 43 Uniform Resource Name (URN) 4
Sigfox 29 UNIS 79
SigTuple 108 Unsupervised learning 92
Simple Authentication And Security User Self Provisioning 68
Layer (SASL) 33 User-​centric edge-​based 75
Smart Grids 6
SMARTIE 14 Vehicular Ad Hoc Networks 6
SmartSantander 78
Stanford NLP 90 WiFi 27
Static Resource Provisioning 68 Wipro Holmes 95
Stochastic Gradient Descent 101 Wireless Sensor Networks (WSN)
Summarization 91 3
Supervised learning 92
Support vector machines (SVM) Zeroth Cognitive Computing
104 Platform 88
System Level offloading 66 Zigbee 27