See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/341817285

Prevent DDOS Attack in Cloud Using Machine Learning

Article · June 2016

CITATION READS

1 289

2 authors:

Anku Jaiswal Chidananda Murthy

Pulchowk Campus(IOE) Jain University
8 PUBLICATIONS   16 CITATIONS    7 PUBLICATIONS   23 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

1.Pneumonial Detection using CNN View project

Maters thesis View project

All content following this page was uploaded by Anku Jaiswal on 02 June 2020.

The user has requested enhancement of the downloaded file.

 Volume 6, Issue 6, June 2016 ISSN: 2277 128X
International Journal of Advanced Research in
Computer Science and Software Engineering
Research Paper
Available online at: www.ijarcsse.com
Prevent DDOS Attack in Cloud Using Machine Learning
Anku Jaiswal Chidananda Murthy P Madhu BR
M.Tech Student, Assistant Professor, Assistant Professor,
SET, Jain University, SET, Jain University, SET, Jain University,
Bengaluru, India Bengaluru, India Bengaluru, India

Abstract: Cloud computing being one of the most demanding technologies has gained a lot of fame due to various
services it provides in the form of utility model. But as we know each technology has both advantages and
disadvantages. One of the biggest disadvantages of cloud is security issue. There are many threats to the cloud
security and Distributed Denial of Service (DDOS) attack is one of the severe threats. DDOS is a simple but a very
powerful attack which makes the resources unavailable to the legitimate client. It is very hard to prevent this type of
attack because of different types in which hacker may attack. Distinguishing between malicious and legitimate traffic
is a tedious work and also manually filtering the traffic is not possible because of large number of hosts. Resource
exhaustion is one of the most common attacks seen today and one of the best techniques would be to build intelligence
to the machine so that it could proactively make decision and distinguish between good and bad traffic automatically.
This is possible with machine learning techniques. In this paper we have discussed how machine learning can help us
to prevent DDOS attack.

Keywords: Cloud Computing, DDOS attack, Machine Learning, Artificial Intelligence

I. INTRODUCTION
With increase in usage of cloud computing to provide the computing resources on demand over the internet, so is the
need to secure the computing resources to provide the reliable and secure services to its user by maintaining the
confidentiality, integrity and authenticity constraints. Since cloud uses the internet to provide the services, it has become
much vulnerable for the various types of attack. One of the major attack issues faced by the cloud platform is DDOS
attack. It is a special type of DOS attack in which attacker uses the network of infected computers which are actually bots
to exhaust the resources of the target system. The DDOS attack can be mainly classified in to two categories: Bandwidth
depletion and resource depletion. The attacks such as UDP and ICMP come under bandwidth depletion and the attack
such as TCP sync and IP flood attack comes under resource depletion. It is one of the very notorious types of attack
which can make the service unavailable to the legitimate user for very long time hampering the credibility and reliability
of the cloud service providers. Most of the defense techniques developed so far has focused on monitoring the network
layers to prevent the attack, but the attacker has now shifted their concentration to the application layer since it performs
more number of computations per packet so the probability of resource exhaustion increases in this layers.

II. RELATED WORK

Cloud computing one of the most used technologies is still in its stage of development as it suffers different types of
security issues. Hence the trust between the client and provider is compromised. To solve these issues a number of
solutions based on cryptography, data portioning, multi agent and machine learning can be proposed. [1].According to a
survey, 1200 approx. DDOS attack occurs daily. Thousands of DDOS attack can be classified on the basis of Bandwidth
depletion and resource depletion. There are various attacks such as Flood attack, Amplification attack, TCP SYN attack.
One of the solution to these attack is collaborative use of various solutions such as Firewall, Intrusion detection, Content
Filtering and VPN (Virtual Private Networks)[2]. The impact of flooding DDOS attack in cloud can be analyzed by using
cloudsim.
This can be done by simulating DDOS attack on the VM instances under Eucalyptus and hence the computational
time of VM under attack can be observed [3]. DDOS attack is a very powerful and notorious attack to cloud computing.
Defending against this attack is not an easy task as the attacker may use different way to affect the system.
Differentiating legitimate an d illegitimate traffic is a tedious task and hence preventing the attack due to various factors
such as: Multiple source, filter placement, throughput. Hence automatic defense technique using machine learning
technique can be proposed to explore the potential of Artificial Neural Network which helps in preventing DDOS attack
[4]. Although a large number of defense techniques has been proposed, preventing DDOS attack is a tough task because
today most of the defense techniques rely on human analysts to differentiate between legitimate and illegitimate traffic.
Many human relying techniques such as Signature detection , anomaly detection takes up to hours and has many flaws
which makes the attack detection a tedious task . Machine learning techniques can be used to make autonomous IDS
techniques to prevent DDOS attack [5].Support vector machine is one of the most common method used in machine
learning techniques. In this method a training example is taken and by using SVM a model can be used to predict

© 2016, IJARCSSE All Rights Reserved Page | 575

 Jaiswal et al., International Journal of Advanced Research in Computer Science and Software Engineering 6(6),
June- 2016, pp. 575-579
whether the new example falls into the category. Hence new types nonlinear problems can be solved by using SVM
[6].Decision tree is another type of machine learning algorithm which uses previous data sets to find new type of data set.
Decision tree can be used in intrusion detection system. One of the advantage of this algorithm is it works well with huge
data set. Decision tree can be used to find new types of attacks [7].Different types of Machine learning techniques are
used to find the detection of DDOS attack. But each technique has its own advantages and disadvantages. Hence various
techniques are compared and studied to find an effective Intrusion Detection System [8].

III. DDOS ATTACK

In distributed denial of service attacks, the attackers uses the multiple system to flood the resources and bandwidth of
the web server or any target systems, to make the services or resources provided by them unavailable for indefinite time
or temporarily for the intended users. The attacks are mainly generated by the multiple system compromised by the
botnet to flood the target system. DDOS attack composed of four elements mainly:
 The attack is initiated by the real attacker.
 All the compromised hosts which have bots running on them are controlled by the Master.
 All the packets which results in flooding of the target resources are generated by the zombies.
 The target system mainly web servers are the victims.

Attackers are always one step ahead the detector. A lot of tool has been developed to solve the problem of network
layer. Hence the hackers are shifting from network layer to application layer and their main aim is resource exhaustion as
it requires less traffic.

Figure 1: DDOS attack architecture

A. Types of DDOS Attack

Table I Types of DDOS
NAME DESCRIPTION
Smurf Attack This attack work with flooding the victim’s bandwidth.
Ping Flood and Ping of Death This attack is similar to smurf attack but the victim is flooded with thousands of
ping packet.
TCP SYN flood This attack aims at exploiting the server CPU memory.
UDP flood In this attack UDP packets are flooded against server.
HTTP Flood Attacks Http floods, which enable attackers to POST large amounts of data to
applications
Teardrop attack This attack utilized vulnerability in the earlier Microsoft
Operating Systems and some older versions of Linux whereby improperly
framed or overlapping IP fragments were sent to the victim thereby crashing it
Infrastructure attacks Infrastructure attacks target the Internet’s infrastructure
Such as the DNS root servers. They can be extremely dangerous as they have
the Potential to bring down the entire Internet.

B. Recent Technique And Their Disadvantages

Table III Techniques and Disadvantages
NAME DEFINITION DISADVANTAGE
Blackholing the process of a service provider blocking all Legitimate packets are discarded along with malicious
illegitimate traffic and sending the diverted attack traffic.
traffic to a “black hole” where it is discarded.
Routers use access control lists (ACLs) to filter out DDOS attacks generally use valid protocols that are

© 2016, IJARCSSE All Rights Reserved Page | 576

 Jaiswal et al., International Journal of Advanced Research in Computer Science and Software Engineering 6(6),
June- 2016, pp. 575-579
“undesirable” traffic, defend against DDOS essential for an Internet presence, rendering protocol
attacks filtering a less effective defence
Firewalls Blocks certain sites They are not purpose-built DDOS prevention devices
lack of anomaly detection location
IDS excellent application layer attack-detection They cannot detect DDOS attacks
capabilities using valid packets—and most of today’s attacks use
valid packets
Signature involves searching network traffic for a They only detect known attacks, a signature must be
detection series of bytes or packet sequences known to created for every attack, and novel attacks cannot be
be malicious detected.
Anomaly The anomaly detection technique centres on Disadvantage of anomaly-detection engines is the
detection the concept of a baseline for network difficultly of defining rules.
behaviour. This baseline is a description of
accepted network behaviour, which is
learned or specified by the network
administrators, or both.

IV. DIFFICULTIES IN PREVENTING DDOS ATTACK

Sites don’t know where the attacks are coming from
• Firewalls aren’t designed to handle DDOS attacks
• The defense can’t be mounted on the hosting provider’s infrastructure
• The tools are much more dynamic now. Humans and human processes are not fast enough to keep up with the
change.
• It is very difficult to determine the origin of the attack.
• Since the attacks are generated from multiple sources it is very diificult to prevent the attack.

V. MACHINE LEARNING
Machine learning is a type of artificial intelligence (AI) that provides computers with the ability to learn without
being explicitly programmed. Machine learning focuses on the development of computer programs that can teach
themselves to grow and change when exposed to new data.
Machine learning techniques have ability to implement a system that can learn from data. For example, a machine
learning system could be trained on incoming packets to learn to distinguish between intrusive and normal packet. After
learning, it can then be used to classify new incoming packets into intrusive and normal packets.

Fig 2: Classification of Machine Learning

A. Artificial neural network

The artificial neural network(ANN) draws its inspiration from the biological neural network which can be used to
estimate the given fuction from the given unknown inputs. As the human brain, the ANN is too composed of multiple
nodes and connected by the link and each link is associated with some weight value, each nodes take the input and
produces the output, which is again passed as the input to the other nodes.
The ANN can be used to determine the complex relationship and pattern between the input data and is considered
very effective data modeling tool.

© 2016, IJARCSSE All Rights Reserved Page | 577

 Jaiswal et al., International Journal of Advanced Research in Computer Science and Software Engineering 6(6),
June- 2016, pp. 575-579

Fig 3: Artificial Neural Network

VI. DDOS PREVENTION APPROACHES

According to [4], machine learning[9][10] can be used to prevent DDOS attack in an automated fashion. Our
proposed system has following features:
• Fully automated system to prevent DDOS attack.
• Focus on resource utilization rather than packet monitoring.
• Using Artificial Neural Network for attack detection and storing its result in sample database for future
reference.

The proposed system consists of following steps:

• A load monitoring node which continuously monitor the system resources (CPU, NETWORK). If the utilized
resource is more than the given threshold value, the incoming packets is identified as anomalous.
• A traffic monitoring node which continuously monitor the packets from different network layers. As packets
from different layers are analyzed, so we have sufficient amount of data to analyze whether the request is
legitimate or illegitimate.
• Now the server which is the central component of the system, it performs two basic tasks: it receives data from
the traffic and load monitoring system and stores it for further processing. Secondly, it continuously monitors
the system. If the system is found to be operating in normal condition then baseline profile for these request is
maintained and if the system is found under the attack then the feature extracted from the load and traffic
monitoring system is store into the Sample Database.

Fig 4: Architecture of the system using neural network

© 2016, IJARCSSE All Rights Reserved Page | 578
 Jaiswal et al., International Journal of Advanced Research in Computer Science and Software Engineering 6(6),
June- 2016, pp. 575-579
• Database stores the baseline profile data and the data extracted by load and traffic monitoring systems which is
used as input for training algorithm.
• Now ANN uses the baseline profile data to train the algorithm and distinguish the difference between the normal
packet and the malicious packet which can be used to prevent the attack.
• The filtering node uses the algorithm produced by ANN to filter and drop the malicious packets.
• The monitoring node is basically a web application which helps the management to continuously manage the
state of the system.

VII. CONCLUSION
DDOS attack detection is a very complex and complicated problem for cloud computing technology. Despite of using
various techniques, DDOS attack is one of the most vulnerable attack .Machine learning based on artificial neural
network can be used to achieve excellent solution as it uses automated system. Rather than analyzing network traffic,
main aim of this technique is to monitor resources which make it an efficient technique.

VIII. FUTURE WORK

DDOS attack has become major security issue for the cloud computing platform. A lot of research work is required to
make the system more secure and provide the trusted and reliable services to its user. Since the system is under the
continuous attack, the existing security measures can be improved with the machine learning technique for automated
security system, so that very little human intervention is required for monitoring, detecting and preventing the attack.

REFERENCES
[1] Y. Ghebghoub, S. Oukid, and O. Boussaid; “A Survey on Security Issues and the Existing Solutions in Cloud
Computing”, International Journal of Computer and Electrical Engineering, Vol. 5, No. 6, December 2013
[2] Raj Kumar P. Arun,S. Selvakumar; “Distributed Denial of service threat in collaborative environment- A survey
on DDOS tools and Traceback mechanism”, IEEE International Advance Computing Conference,2009
[3] Mr S.Karthik, Prof J.J.Shah; “Analysis of Simulation of DDOS Attack in Cloud” , Information Communication
and Embedded Systems (ICICES), 2014 International Conference
[4] Stefan Seufert and Darragh O’Brien ,”Machine Learning for Automatic Defence against Distributed Denial of
Service Attacks”, ICC 2007 proceedings.
[5] Chris Sinclair, Lyn Pierce, Sara Matzner “An Application of Machine Learning to Network Intrusion
Detection”. Phoenix, AZ 06 Dec 1999-10 Dec 1999
[6] J. Burges, "A tutorial on support vector machines for pattern recognition", Data Mining and Knowledge
Discovery, vol. 2, pp. 12 1- 167, 1998.
[7] Kamarularifin Abd Jalil, Muhammad Hilmi Kamarudin, Mohamad Noorman Masrek; “Comparison of Machine
Learning Algorithms Performance in Detecting Network Intrusion”, 2010 International Conference on
Networking and Information Technology
[8] Jayveer Singh , Manisha J. Nene , “A Survey on Machine Learning Techniques for Intrusion Detection
Systems” , International Journal of Advanced Research in Computer and Communication Engineering Vol. 2,
Issue 11, November 2013
[9] Sujay Apale, Rupesh Kamble, Manoj Ghodekar, Hitesh Nemade, Rina Waghmode; “Defense Mechanism For
Ddos Attack Through Machine Learning”, IJRET: International Journal of Research in Engineering and
Technology eISSN: 2319-1163 | pISSN: 2321-7308
[10] Sergio Armando Guti´errez, John Willian Branch Grupo GIDIA, “Application of Machine Learning Techniques
to Distributed Denial of Service (DDoS) Attack Detection: A Systematic Literature Review”

© 2016, IJARCSSE All Rights Reserved Page | 579

View publication stats