IRJET- Improving Cyber Security using Artificial Intelligence

INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056
VOLUME: 06 ISSUE: 05 | MAY 2019 WWW.IRJET.NET P-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 6172
Improving Cyber Security using Artificial Intelligence
J Ramyashree1 and Hemavathy R2
1,2Department of Computer Science and Engineering, Rashtreeya Vidyalaya College of Engineering, Bangalore,
Karnataka, India.
----------------------------------------------------------------------------***--------------------------------------------------------------------------
Abstract — In a world where there is an increasing
number of cyber security threats, there is a need for faster
and more cost-effective way of detecting and responding
to threats. This paper deals with how we can use artificial
intelligence to improve cyber security. The methodology
used is to use Splunk to get the fields which we are
interested in and feed that data to K-means algorithm
which forms clusters of subsets of the attacks. Then those
clusters are fed into individual ANNs from which the
aggregated results are fed into SVM which outputs
whether the attack is malicious, non-malicious or benign.
Keywords — Artificial Intelligence, Cyber Security,
Machine Learning, Supervised Learning.
I.INTRODUCTION
With the rise of Machine Learning, there is also a rise in
cyber-criminality. Due to this we need a faster way of
detecting threats and responding properly. Using ML
makes the task faster. What takes a Cyber Security
operation team two days to figure out, ML can figure it out
in a day and since ML continuously learns based on
experience and results, it will take lesser and lesser time to
understand and figure out what’s happening. Machine
learning is technique wherein the machine, given the
necessary data will be trained and thereon will start
learning on its own. Its accuracy of prediction increases
with more data and time. Three main techniques for
machine learning are supervised, semi-supervised and
unsupervised. Supervised learning refers to a technique of
training on the data which is already labeled and
predicting output from the given labeled input.
Unsupervised learning refers to a technique where the
algorithm tries to find patterns in the data and gives
output. Semi-Supervised technique is where the algorithm
trains on a labelled and unlabeled but mainly on unlabeled
data. K- Means clustering is an unsupervised technique. It
is a technique where we initialize k points randomly, each
point is categorized to its closest mean, and this step is
repeated till we get our clusters. ANN is artificial neural
network where the neural nets are trained by giving
weights to yield various results. SVM is a supervised
technique wherein there is a line called hyperplane which
classifies the plotted data points into the different clusters.
There are many kinds of cyber threats such as
ransomware, botnet, trojan, Advanced persistent
threat, DDOS, Man in the middle attack, phishing,
spyware, wiper attacks, malware, etc. An Intrusion
detection system (IDS) is a process wherein the whole
intrusion detection process is automated. An Intrusion
prevention system (IPS) is a process wherein it can not
only detect attacks but also prevent the possible
attacks. Given the increasing number of cyber-crimes
today, we need a system which can detect and prevent
attacks.
The attack which occurs could be accidental,intentional
or unintentional and cause various issues like loss of
data, data access, data modification, etc.
The whole purpose of computer security is to detect
and prevent use of the system without any
authorization. Destruction or any kind of modification
to data is prevented.
Fire-walls, anti-viruses, suites are some of the examples
of computer-security. There are different kinds of
Threats and classified such as-
A. Physical Threats
Physical threats when there is physical damage to the
hardware and infrastructure caused by fire,
earthquake, landslide, etc
B. Non-physical Threats
These are the kind of attacks which could lead loss or
corruption of sensitive data, monitoring of computer
systems illegally, breaches in cyber security, etc. Some
of the examples of these kind of threats are key loggers,
virus, worms, adware, phishing, DDOS, spyware, DOS,
etc.
II. LITERATURE REVIEW
A brief literature survey was carried out to gain
understanding about various concepts like cyber
security threats and machine learning algorithms.
Excerpts from a few of the papers/ journal referred are
as follow:

Noor Ahmed et. al. gives an overall description about
the various kinds of IDS and how different algorithms
are used to detect IDS and their performance. [1].
Jack W Stokes et. al speaks about how static malware
analysis is not so efficient due to the fact that malware is
usually in the encrypted format and how attackers can
make small changes to the input and avoid the malware
being detected. The technique used in this paper is
adversarial attack strategies against a dynamic malware
analysis classification system. [2].
Cristian Pascariu et. al speak about applying machine
learning for analysing the dynamic malware process.
This is applied for Windows operating System. This
technique focuses on the process hierarchy in a way
that that will capture information about the created
process and its parent process. The output will have a
higher score if its malicious and a lower score if it’s safe
activity.[3].
Rahul Vigneswaran K et. al explains how to exploit the
randomness of the incoming attacks which is not
visible to the human eye but can be made better by
using Artificial intelligence. Hence, by training the
algorithm with existing cyber attacks, prediction will be
done whether the attack is malicious or not. Due to this,
the time and cost to detect becomes increasing less.[4].
S. Ponmaniraj et. al takes the TCP/IP parameters as
input and analysing the network security. The intrusion
detection system deals with huge networks congestion
highly irregular based data pattern. The difference
between regular and irregular or anomaly pattern data
sets are, regular pattern has certain static derived input
and expected output sets whereas irregular all kind of
data are processing as input and dynamically they are
getting change in its behaviours. Regular data set
cannot change over a time whereas irregular pattern
will get change and therefore classification and
prediction model must be changed accordingly. In
networks intrusion detection can be classified into
misuse and anomaly detection. Detection of misuse
found by trained data set and well known attacking
methods with predicted output results. Anomaly
detection dealing with unknown data sets which shows
variations in usual set of normal behaviours or patterns
unknown to the system to detect. [5].
III. METHODS
A. Experimental Setup
The various tools and programming languages used while
carrying out the experiment is as follows:
 Splunk Enterprise version
 Jupyter Editor
 Splunk Programming Language
 Python
B. Design Modules
The modules that make up the entire system is depicted in
Figure 1.
Figure 1. Design Modules
Splunk is a software that captures real time data and
indexes,correlates and presents the data in a way which
can be searched and visualized easily which can be widely
and esily applied for cyber security.
K-means is an algorithm which is unsupervised
learning technique which is used to find patterns in the
data and find clusters. Data points based on the
similarity of data points are clustered. The number of
clusters are pre-determined and that is known as K.
Artificial neural networks are statistical models which try
to model the human brain. There are nodes in the ANN
which conveys the data to the next neuron and so on till it
is fed to the output nodes. After this, labels can be given to
the output.
SVM is an algorithm which is used primarily for
classification and it is a type of supervised type of
algorithm. There is a hyper plane which is a line which
classifies the labelled data points into different classes.
C. Implementation
The data is fed from the NSL-KDD dataset into the Splunk
Indexer which is used for parsing and indexing the data,
the resulting data is fed into Splunk Forwarder which is
used for data forwarding. The reason Splunk is chosen for
choosing the relevant features instead of PCA is because

the types of malicious activities is fixed and hence the
features will be constant and the efficiency will be more.
Based on the features selected, the data is split into 7:3
ratio of training and testing respectively. The training data
is fed into the K-Means algorithm and clusters of data are
formed.
This clusters of data are fed into the individual ANN and
the various results are obtained.
The aggregated results are fed into the SVM which gives
the results of whether the attack is malicious, non
malicious or benign.
IV. RESULTS AND ANALYSIS
The data that are analyzed using these algorithms. All the
algorithms were trained with a total of 400 samples with
280 training samples and tested with test data set
containing 120 samples. All the four different types of
algorithms were compared and the accuracies obtained
were:
Method Accuracy Precision Recall F1 Score
K
Means
and
SVM
0.7843 0.81 0.83 0.79
PCA, K
Means
and
SVM
0.8540 0.86 0.84 0.82
Splunk,
K-
Means
and
SVM
0.8756 0.84 0.86 0.87
Splunk,
K-
Means,
ANN
and
SVM
0.9256 0.94 0.93 0.94
Table I : Comparison of different models
By this we can observe that by far, using the
combination of algorithms initially it could be time
consuming but as training progresses, this is by far the
most efficient method.
V. CONCLUSION
It is impossible to prevent cyber-crimes completely, by
using the current technologies in security. In this paper,
an efficient method intrusion detection system is
proposed to detect whether the attack is malicious,
non-malicious or benign. Taking the help of Splunk to
get the important features after which the data is fed to
K Means to get a cluster of subsets.After obtaining the
clusters, training will be given to the individual
artificial neural networks separately and after which
the aggregated results is fed into the SVM which
classifies the attacks as malicious, non-malicious or
benign. The dataset considered here is the NSL-KDD
dataset. The evaluation metrics here is accuracy,
precision, recall and F-square. The results show that
the approached method is efficient.
REFERENCES
[1] Noor Ahmed Biswas and Wasima Matin Tammi “
FP-ANK: An Improvised Intrusion Detection System
with Hybridization of Neural Network and K-Means
Clustering over Feature Selection by PCA ” in 18th
Internation Conference on Computer and Information
Technology (ICCIT),2015.
[2] Jack W Stokes and Md Amran Siddiqui “Detecting
Cyber Attacks with explanations and Expert Feedback”,
in: ICASSP 2019 - 2019 IEEE International Conference
on Acoustics, Speech and Signal Processing (ICASSP)
[3] Cristian Pascariu and Ionut-Daniel Barbu “Dynamic
analysis of malware using artificial neural networks” in
ECAI 2017 - International Conference – 9th
Edition,2018,pp.67-71.
[4] Rahul Vigneswaran K, Vinayakumar R “ Evaluating
Shallow and Deep Neural Networks for Network
Intrusion Detection Systems in Cyber Security” in 9th
ICCCNT 2018,pp.26-29.
[5] K. Ibrahimi and M. Ouaddane, “Management of
intrusion detection systems based-kdd99: Analysis
with lda and pca,” in Wireless Networks and Mobile
Communications (WINCOM), 2017 International
Conference on. IEEE, 2017, pp. 1–6.
[6] N. Moustafa and J. Slay, “The significant features of
the unsw-nb15 and the kdd99 data sets for network

intrusion detection systems,” in Building Analysis
Datasets and Gathering Experience Returns for Security
(BADGERS), 2015 4th International Workshop on.
IEEE, 2015, pp. 25–31.
[7] L. Sun, T. Anthony, H. Z. Xia, J. Chen, X. Huang, and Y.
Zhang, “Detection and classification of malicious
patterns in network traffic using benford’s law,” in
Asia-Pacific Signal and Information Processing
Association Annual Summit and Conference (APSIPA
ASC), 2017. IEEE, 2017, pp. 864–872.
[8] S. M. Almansob and S. S. Lomte, “Addressing
challenges for intrusion detection system using naive
bayes and pca algorithm,” in Convergence in
Technology (I2CT), 2017 2nd International Conference
for. IEEE, 2017, pp. 565–568.
[9] M. C. Raja and M. M. A. Rabbani, “Combined analysis
of support vector machine and principle component
analysis for ids,” in IEEE International Conference on
Communication and Electronics Systems, 2016, pp. 1–5.
[10] S. Aljawarneh, M. Aldwairi, and M. B. Yassein,
“Anomaly-based intrusion detection system through
feature selection analysis and building hybrid efficient
model,” Journal of Computational Science, vol. 25, pp.
152–160, 2018.
[11] K. Graves, Ceh: Official certified ethical hacker
review guide: Exam 312-50. John Wiley & Sons, 2007.
[12] R. Christopher, “Port scanning techniques and the
defense against them,” SANS Institute, 2001.
[13] S. Staniford, J. A. Hoagland, and J. M. McAlerney,
“Practical automated detection of stealthy portscans,”
Journal of Computer Security, vol. 10, no. 1-2, pp. 105–
136, 2002.
[14] E. Gerhards-padilla and F. Fkie, “Intrusion
Detection in Tactical Multi-Hop Networks,” 2009.
[15] S. Robertson, E. V. Siegel, M. Miller, and S. J. Stolfo,
“Surveillance detection in high bandwidth
environments,” in DARPA Information Survivability
Conference and Exposition, 2003. Proceedings, vol. 1.
IEEE, 2003, pp. 130–138.

IRJET- Improving Cyber Security using Artificial Intelligence

Recommended

More Related Content

What's hot (15)

Similar to IRJET- Improving Cyber Security using Artificial Intelligence (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Improving Cyber Security using Artificial Intelligence