EJERS, European Journal of Engineering Research and Science

Vol. 5, No. 10, October 2020

Smart Intrusion Detection System Comprised of Machine

Learning and Deep Learning
Shah Md. Istiaque, Asif Iqbal Khan, Sajjad Waheed

security since aggressors utilize fresher methods for

Abstract — In the present world, digital intruders can exploit ambushing the system just as its clients [3].
the vulnerabilities of a network and are capable to collapse even
a country. Attack in Estonia by digital intruders, attack in Iran's Over the years, operating system security technology has been
nuclear plant and intrusion of spyware in smart phone depicts
the efficiency of attackers. Furthermore, centralized firewall
upgraded to forestall the issues of confidentiality, integrity,
system is not enough for ensuring a secured network. Hence, in and availability of a network [4]. Intrusion Detection System
the age of big data, where availability of data is huge and (IDS) is a prudent early warning system for a network. For the
computation capability of PC is also high, there machine most part, IDS alarms the client before the system gear is
learning and network security have become two inseparable imperiled when it identifies inner and outside intrusion [5].
issues. At first, network manager executes IDS physically by
In this thesis, KDD Cup’99 intrusion detection dataset is used.
Total 3, 11,030 numbers of records with 41 features are available
observing the system through a console [6]. However, the
in the dataset. For finding the anomalies of the network four objective of this research is to propose a system security
machine learning methods are used like Classification and framework actualizing the Artificial Neural Network (ANN),
Regression Tree (CART), Random Forest, Naive Bayes and utilizing Back Propagation (BP) calculation. The
Multi-Layer Perception. Initially all 41 features are used to find examination further breaks down oddity discovery, in light of
out the accuracy. Among all the methods, Random Forest a few AI methods, using different training and testing datasets
provides 98.547% accuracy in intrusion detection which is
maximum, and CART shows maximum accuracy (99.086%) to
[7].
find normal flow of data. Gradually selective 15 features were The remaining part of the paper is developed in the
taken to test the accuracy and it was found that Random Forest following way. Segment 2 provides a literature review
is still efficient (accuracy 98.266%) in detecting the fault of the regarding recent updates, Segment 3 highlights some
network. In both cases MLP found to be a stable method where classification algorithm utilized in the work, which is CART,
accuracy regarding benign data and intrusion are always close Random Forest, Naïve Bayes and MLP. Experimental results
to 95% (93.387%, 94.312% and 95.0075, 93.652% respectively).
Finally, an IDS model is proposed where Random Forest of
and analysis, the core contribution of this paper, is discussed
ML method and MLP of DL method is incorporated, to handle in Segment 4. Lastly, Segment 5 conclude the paper.
the intrusion in a most efficient manner.

Index Terms — Intrusion Detection System, CART, Random II. RELATED WORK
Forest, Naive Bayes, Back-Propagation based MLP. A lot of studies have already been carried out about various
machine learning and deep learning methods and its
application in different fields. Same is not yet exhaustively
I. INTRODUCTION
done in the field of information security. The gap found in the
The Internet has become an important media of regular field of IDS was studied with available resources till today.
correspondence through online media collaboration, email, e- The unique idea of applying the machine learning and deep
learning, and so forth. Additionally, little and large learning methods in the IDS is a new theme in the
organizations have expanded their purchaser base by giving contemporary research arena.
direct client showcasing, web shopping and inter organization
correspondence utilizing essential web correspondence. With A. Study Regarding IDS
the gigantic development of computer network, the total According to each of the detection approaches, IDS
system experiences security weaknesses which are frameworks are separated into two classifications, which are
troublesome and exorbitant to be solved by manufactures [1]. anomaly-based detection and misuse based detection [8], [9].
A few dangers are brought through utilization of incapable Misuse-based IDS can recognize known assaults effectively
and wasteful security instruments welcoming intrusions from yet neglects to discover new assaults which fail to embody
internet hackers [2]. In this way, it is clear that the prevention the rules in the database [10]. In this manner, a database must
technologies set up like malware evacuation programs, be persistently refreshed to store the signature of each assault
antivirus projects and firewalls, neglect to give outright that is known. This IDS type is clearly incapable to identify
new attacks except it is trained [11]. Anomaly based IDS can

Published on October 8, 2020. Sajjad Waheed, Professor Dr., Mawlana Bhashani Science and
Shah Md. Istiaque, Bangladesh University of Professionals, Bangladesh. Technology University, Bangladesh.
(e-mail: [email protected])
Asif Iqbal Khan, Mawlana Bhashani Science and Technology University,
Bangladesh.

DOI: http://dx.doi.org/10.24018/ejers.2020.5.10.2128 Vol 5 | Issue 10 | October 2020 1

 EJERS, European Journal of Engineering Research and Science
Vol. 5, No. 10, October 2020

assemble a typical conduct model and recognizes any detection. Shilpaet Al [26] used fundamental element
significant deviations from the model similar to an evaluation on NSLKDD dataset for feature selection and
interruption. This IDS type can identify new assaults or dimension pruning approach for evaluation on anomaly
obscure one. However, it includes a high pace of false alarms detection. In general, network intrusion detection has been
[12], [13]. broadly improved by applying data mining and machine
learning technique, which has largely utilized individual
B. Study Regarding Dataset
conduct patterns from the community site visitors’ data.
The most significant challenge in assault identification Support Vector Machine (SVM) is used, as a method in a
framework is whether to produce genuine system traffic or to study, to evaluate IDS [27]. Among various approaches of
utilize the accessible benchmark datasets. There is criticism IDS, SVM acts as a classifier with false alarm and detection
about the use of datasets acquired from genuine system traffic rate as a measure of performance. Authors in a study [28] used
as it makes greater uncertainty and there is no such Markov Chain implementation as classifier and Apriori
methodology that obviously discloses how to precisely algorithm to remove isolated data from the database and also
separate between ordinary system traffic and attack traffic. used to judge the performance of NIDS. K-Means, an
This is the explanation behind utilizing the benchmark unsupervised algorithm, is used for classification, defines an
datasets for executing different attack discovery framework unlabeled class to which the clustering is performed.
of this paper. The available attack datasets [14]-[17] are
DARPA 1998, KDD Cup99, NSL KDD, UNSW NB15, etc.
The DARPA 1998, KDD Cup99, and NSL KDD consists of III. PRELIMINARIES
42 attributes including the class label. The UNSW NB15
This section provides a brief background about network
dataset consists of 48 attributes including the class label.
intrusion, and the four intelligent algorithm used in this study.
C. Review Regarding Detection
A. Concept of Network Intrusion
Multiple detection methods have been carried out in
Modern technology has broken the border of digital
various literatures. It includes traditional detection, ML-
intrusion and also digital threat. Attack in Estonia, Iran’s
based and DL Neural-network based detection. In few
nuclear power plant, digital espionage, financial damage-all
research hybrid method is also used. Various detection
of these are the newest threat of modern internet technology.
techniques are analyzed in the following discussion.
Digital intrusion is the first step and the most common type
D. Traditional Detection of attack or threat [29]. Then onward malwares are injected
A sandbox, in computer security, is a security component or further important arsenals are used. Thereby, if intrusions
wherein a different, confined condition is made and in which are monitored and checked then first line defense can possibly
several functions are restricted [18]. A sandbox is regularly be achieved.
utilized when untested code or entrusted programs from
outsider sources are being utilized. Sandbox also has few
1. Reconnaissance
constrain. Some sandbox apparatuses just deal with explicit
sorts of PDF assaults like MD Scan for Java Script, [19]
Nozzle for heap spraying [20], or it only records dynamic 7. Obfuscation 2. Scanning and
and Exfiltration Weaponization
behavior of a system and still requires manual analysis to
detect as in the case of CW Sandbox [21].
Huaibin Wang, HaiyunZhou, Chundong Wanghas
discussed about VM-based different IDSs [22]. They have
recommended to deploy VM-based numerous IDSs in each
layer to observe specific virtual component. Additionally, 6. Action on 3. Access and
they have also proposed the cloud alliance view, by the objective Escalation

communication agents exchanging shared cautions

commonly to withstand Denial of Service (DoS) and
Distributed Denial-of-Service (DDoS). On this premise, they
have accomplished an identity authentication of the 5. Installation
and Command 4. Sustainment
communication agents, to improve the unwavering quality of Control
the alarms. Through the evaluation of simulation results, the
Fig. 1: Network Attack Cycle.
proposed device framework had a benefit for observing VMs
on the detection rate.
B. Concept of Network Attack
E. Machine Learning (ML) and Deep Learning (DL) Based Analyzing the Fig. 1 it is clearly understood that first three
Detection
steps of network attack cycle are related to intrusion.
Machine learning algorithm learns from data [23]. Tom Therefore, it plays a vital role in overall attack cycle.
Mitchell precisely defines it as a computer program which
learns from experience in respect to task and final outcome C. Overview Regarding Algorithm
is the performance [24]. There are many types of algorithms practiced in machine
Vipin Kumar [25] used k mean clustering approach on learning. But four suitable methods are utilized for
NSLKDD dataset to perceive the accuracy for intrusion performance analysis.

DOI: http://dx.doi.org/10.24018/ejers.2020.5.10.2128 Vol 5 | Issue 10 | October 2020 2

 EJERS, European Journal of Engineering Research and Science
Vol. 5, No. 10, October 2020

D. Classification and Regression Tree means one particular feature does not affect the other.
Leo Bremen introduced the term CART. CART refers Therefore, this technique is called naïve [32].
Decision Tree algorithm. It is used for classification or G. Multi-Layer Perception
regression predictive modeling problems. Classically, this
Multi-layer perceptron is a deep learning technique where
algorithm is mentioned as “decision trees”. However, they
more than one linear layer (combination of neurons) is
are also mentioned by the more modern term CART, on some
involved. In a three-layered network, first layer will be
platforms like R [30].
the input layer and last one will be output layer and with a
hidden layer in between [33].
Height > 175cm

Yes No
Weight >75kg
Input Output
Male Yes No

Male Female
Input First Second Output
Layer Hidden Output Layer
Fig. 2: Classification and Regression Tree. Layer Layer

Fig. 4. Multi-Layer Perception Model (MLP).

The tree may be preserved to file as a graph or a set of rules.
For example, it is given below as a set of rules.
IV. EXPERIMENTAL RESULTS AND ANALYSIS
• If Height > 175 cm, Then Male.
• If Height <= 175 cm AND Weight >75 kg, Then Male. This section discusses the performance of the proposed IDS
• If Height <= 175 cm AND Weight <= 75 kg, Then model which is furthermore analyzed with two experiments.
Female. In case of MLP, all the features are supplied as input to the
• Make Predictions with CART Models. neural network and then are trained using back-propagation.
Then the accuracy is calculated from the test data. But in case
E. Random Forest of Random Forest, CART, and Naive Bayes, necessary
Random forest consists of a large number of individual parameters are utilized. However, in the second experiment,
decision trees. It operates as an ensemble. Each individual tree only the most important features are extracted, and then
splits out a class prediction. The class with the most votes supplied as input along with the training dataset. Finally, the
become model’s prediction (see figure below) [31]. results obtained are presented in tabular as well as graphical
form.
A. Application of ML Methods with Generalized Features
In the first experiment four methods are applied to find out
accuracy in both normal flow of data and also for intrusion.
Among all the ML Method, CART and MLP has provided a
better accuracy. Whereas Random Forest has provided a
greater intrusion detection which is 98.547%. In Fig. 5
graphical presentation has been displayed the same.

TABLE I: TEST ACCURACY FOR NORMAL FLOW OF DATA AND INTRUSION

DETECTION USING 41 FEATURES
Type of Algorithm Type of Data Accuracy
Normal 85.387
Random Forest
Intrusion 98.547
Normal 99.086
Fig. 3. Random Forest. CART
Intrusion 96.51
F. Naive Bayes Classifier Normal 85.606
Naive Bayes
In the arena of Machine Learning, Naïve Bayes performs Intrusion 93.265
classification. The core essence of the classifier depends on Normal 93.387
Bayes theorem. MLP
Intrusion 94.312
P ( B / A) P ( A)
P( A / B) =
P( B)
This theorem follows probability of happening and occurs
where the later one is evidence and the initial one is
hypothesis. Predictors or features are independent which

DOI: http://dx.doi.org/10.24018/ejers.2020.5.10.2128 Vol 5 | Issue 10 | October 2020 3

 EJERS, European Journal of Engineering Research and Science
Vol. 5, No. 10, October 2020

Type of Algorithm Type of Algorithm

100 100

95 95

90 90
85
85
80
80
75
75

Normal

Normal

Normal

Normal
Intrusion

Intrusion

Intrusion

Intrusion
Normal

Normal

Normal

Normal
Intrusion

Intrusion

Intrusion

Intrusion
Accuracy

Accuracy
Random CART Navie Bayes MLP
Random Forest CART Navie Bayes MLP Forest

Fig. 5. Comparison among the ML and DL Methods for finding out Fig. 6. Comparison among the ML and DL Methods for finding out
Accuracy using Generic Features. Accuracy using Selective Features

B. Application of ML and DL Methods with Selective TABLE III: COMPARISON OF ACCURACY AMONG ML AND DL METHODS
Features USING GENERIC AND SELECTIVE FEATURES
Repeatedly in the second experiment, four methods are Type of Type of Accuracy Generic Accuracy Selective
Algorithm Data Features Features
applied to find out accuracy in both normal flow of data and
Random Normal 85.387 90.903
also for intrusion. But in this case selective features are applied
Forest Intrusion 98.547 98.266
in four different ML Methods like Random Forest, CART,
Normal 99.086 98.246
Naive Bayes, & MLP. Here, CART and Naive Bayes has CART
Intrusion 96.51 93.167
provided a better accuracy although Random Forest has also
Normal 85.606 98.331
provided a better intrusion detection like previous which is Naive Bayes
Intrusion 93.265 93.458
98.266%. Performance of MLP has also been displayed a
Normal 93.387 95.007
significant improvement. In the Fig. 6 graphical presentation MLP
Intrusion 94.312 93.652
has also displayed the overall performance.

TABLE II: TEST ACCURACY FOR NORMAL FLOW OF DATA AND INTRUSION Type of Algorithm
DETECTION USING SELECTIVE 15 FEATURES
105
Type of Algorithm Type of Data Accuracy 100
Normal 90.903 95
Random Forest
Intrusion 98.266 90
Normal 98.246 85
CART
Intrusion 93.167 80
Normal 98.331 75
Normal

Normal

Normal

Normal
Intrusion

Intrusion

Intrusion

Intrusion

Naive Bayes
Intrusion 93.458
Accuracy

Normal 95.007
MLP
Intrusion 93.652 Random CART Naive Bayes MLP
Forest
C. Analytical Review Class Name Accuracy with 41 Feature
Experimental results in both cases have displayed Class Name Accuracy with 15 Features
reasonably good performance. Use of selective features and Fig. 7. Graphical Representation and Comparison of Accuracy among ML
Methods using Generic and Selective Features.
elimination of few less important parameters have also
improved the overall performance. After analyzing overall
results, Classification and Regression Tree is found to be a D. Proposed IDS Model
stable and better method keeping in mind that Random Forest A model consisting ML and DL method is proposed in Fig.
provided the best intrusion detection in both cases. 8. Here MLP with Back Propagation algorithm is used and
Random Forest is taken as ML method. The selected methods
are used considering the performance in accuracy.

DOI: http://dx.doi.org/10.24018/ejers.2020.5.10.2128 Vol 5 | Issue 10 | October 2020 4

 EJERS, European Journal of Engineering Research and Science
Vol. 5, No. 10, October 2020

[5] Olanrewaju R F, Habaebi M H. ''Malicious behavior of node and its

significant security techniques in MANET- A review,'' Australian
Journal of Basic and Applied Sciences.2013;7(12):286−93.
[6] The history of intrusion detection systems (IDS) Part 1 Threat stack.
Date accessed: 09/09/2015.https://www.threatstack.com/blog/the-
history-of-intrusion-detection-systems-ids-part-1/.
[7] M. Shyu, S. Chen, K. Sarinnapakorn, and L. Chang, “A novel anomaly
detection scheme based on principal component classifier,
”Proceedings of the IEEE Foundations and New Directions of Data
Mining Workshop, in conjunction with the Third IEEE International
Conference on Data Mining (ICDM03), pp. 172–179, 2003.
[8] Anderson, J. A. (1995). ''An introduction to Neural Networks,'' MIT
Press.
[9] Rhodes, B. C., Mahaffey, J. A., &Cannady, J. D. (2000). ''Multiple self-
organizing maps for intrusion detection.'' In Proceedings of the
23rdnational information systems security conference (pp. 16-19).
[10] Al-Yaseen, W. L., Othman, Z. A., & Nazri, M. Z. A. (2017). ''Multi-
level hybrid support vector machine and extreme learning machine
based on modified K-means for intrusion detection system.'' Expert
Systems with Applications, 67, 296-303.
[11] Chen, C. M., Chen, Y. L., & Lin, H. C. (2010). ''An efficient network
intrusion detection'', Computer Communications, 33(4), 477-484.
[12] Deepa, A. J., &Kavitha, V. (2012). ''A comprehensive survey on
approaches to intrusion detection system.'' Procedia Engineering,
38,2063-2069.
[13] Thaseen, S., & Kumar, C. A. (2013). ''An analysis of supervised tree
based classifiers for intrusion detection system.'' In Pattern
Recognition, Informatics and Mobile Engineering (PRIME), 2013
International Conference on (pp. 294-299). IEEE.
[14] F. Iglesias, T. Zseby, ''Analysis of network traffic features for anomaly
detection,'' Machine Learning 101 (1-3) (2015) 59–84. doi:10.1007/525
Fig. 8. Proposed IDS Model. s10994-014-5473.
[15] N. Moustafa, J. Slay, ''The evaluation of network anomaly detection
systems: Statistical analysis of the unsw-nb15 data set and the
comparison with the kdd99 data set,'' Information Security Journal: A
V. CONCLUSION Global Perspective 25 (1-3) (2016) 18–31.
Sovereignty of a country is assured by ensuring the border doi:10.1080/19393555.2015.1125974.
[16] M. Tavallaee, E. Bagheri, W. Lu, A. A. Ghorbani, ''A detailed analysis
security. But at the age of twenty first century border security of the kdd cup 99 data set,'' in: Computational Intelligence for Security
is redundant where digital security is not guaranteed as digital and Defense Applications, 2009. CISDA 2009. IEEE Symposium on,
world has got no border. However, living with modern IEEE,2009, pp. 1–6. doi:10.1109/CISDA.2009.5356528.
[17] J. McHugh, ''testing intrusion detection systems: a critique of the 1998
technology is an important arsenal to ensure proper digital 535 and 1999 darpa intrusion detection system evaluations as
security. In this thesis, digital security vulnerabilities are performed by lincoln laboratory”, ACM Transactions on Information
discussed at the first place and subsequently it is manifested and System Security(TISSEC) 3 (4) (2000) 262–294.
doi:10.1145/382912.382923.
that the digital security technologies like malware removal [18] www.techopedia Space issue.
programs, antivirus programs and firewalls, lack success to [19] Z. Tzermias, G. Sykiotakis, M. Polychronakis, and E. P. Markatos,
provide absolute protection. Attackers always employ updated “Combining Static and Dynamic Analysis for the Detection of
Malicious Documents, in Proceeding of the fourth Workshop on
techniques to assault the network, as well as its users. European Workshop on System Security,'' (Salzburg, Austria),2011.
Wherefore, latest DL and ML methods are explored and finally [20] P. Ratanaworabhan, B. Livshits, and B. Zorn, “NOZZLE: A Defense
necessary training and tests are carried out to measure the Against Heap spraying Code Injection Attacks, in SSYM’09
Proceeding soft the 18th conference on USENIX security symposium,''
accuracy of the various DL and ML methods. MLP, one of the (Berkeley, CAUSA), 2009.
deep learning method along with various ML methods like [21] C. Willems, T. Holz, and F. Freiling, “Toward Automated Dynamic
Random Forest, CART, Naïve Bayes is also analyzed. The ML Malware Analysis Using CW Sandbox”.
and DL methods are found to be very prudent in network [22] Huaibin Wang, Haiyun Zhou, ChundongWang “Virtual Machine-
based Intrusion Detection System Framework in Cloud Computing
security. Eventually, an IDS model is proposed where MLP Environment” JCP 2012 Vol.7(10): 2397-2403 ISSN: 1796-203Xdoi:
and Random Forest is put into action. 10.4304/jcp.7.10.2397-2403.
[23] I. Good Fellow, Y. Bengio, and A. Courville, ''Deep Learning,'' The
MIT Press, 2016.
[24] T. Mitchell, ''Machine Learning,'' McGrawHill,1997.
REFERENCES [25] Vipin Kumar, Himadri Chauhan, Dheeraj Panwar, “K-Means
[1] C. E. Land Wehr, A. R. Bull, J. P. McDermott, and W. S. Choi, “A Clustering Approach to Analyze NSL-KDD Intrusion Detection
taxonomy of computer program security flaws”, ACM Comput. Surv. Dataset” International Journal of Soft Computing and Engineering
vol. 26, no .3, pp. 211–254, 1994. (IJSCE)ISSN:2231-2307, Volume-3, Issue-4, September2013.
[2] Olanrewaju R F, Khan B U I, Anwar F, Khan AR, Shaikh FA, Mir MS. [26] Shilpalakhina, Sini Joseph and Bhupendraverma, “Feature
''MANET– A cogitation of its design and security issues, ''Middle-East Reductiousing Principal Component Analysis for Effective Anomaly–
Journal of Scientific Research. 2016;24(10):3094−107. Based Intrusion Detection on NSL-KDD”, International Journal of
[3] KhamphakdeeN, BenjamasN, SaiyodS. ''Network traffic data to ARFF Engineering Science and Technology, Vol.2(6),2010,1790-1799.
converter for association rules technique of data mining,'' IEEE [27] Mohammadpour L, Hussain M, Aryanfar A, Raee VM, Sattar F.
Conference on Open Systems (ICOS), ''Evaluating performance of intrusion detection system using support
IEEE;2014Oct.p.89−93.Crossref. vector machines,'' International Journal of Security and Its
[4] Yu S C, Guo H, Yu G X, Jin X L, Zhang L N, Shao T J. ''The solution Applications. 2015 Sep; 9 (9): 225−34. Cross ref.
to how to select an optimal set of features from many features used to [28] Brindasri S, Saravanan K. ''Evaluation of network intrusion detection
intrusion detection system in wireless sensor network.'' 2010 Second using Markov chain, ''International Journal on Cybernetics and
WRI Global Congress on Intelligent Systems (GCIS), IEEE; 2010 Informatics (IJCI).2014Apr; 3 (2): 11−20. Crossref.
Dec.3.p.368−71.Crossref.

DOI: http://dx.doi.org/10.24018/ejers.2020.5.10.2128 Vol 5 | Issue 10 | October 2020 5

 EJERS, European Journal of Engineering Research and Science
Vol. 5, No. 10, October 2020

[29] Clarence Chio and David Freeman, ''Machine Learning and Security,''
O’REILLY, P.6.
[30] https://machinelearningmastery.com/classification-and-regression-
trees-for-machine-learning/ Accessed on 25 Aug 2020.
[31] https://towardsdatascience.com/ understanding-random-forest-
58381e0602d2Accessed on 25 Aug 2020.
[32] https://towardsdatascience.com/ naive-bayes-classifier-81d512f50a7c.
[33] https://medium.com/@xzz201920/multi-layer-perceptron-mlp-
4e5c020fd28aAccessed on 25 Aug 2020.

Shah Md. Istiaque is a military communication

expert. As part of his career requirement, he
completed graduation from the Department of
Electrical Electronics and Communication
Engineering from Military Institute of Science and
Technology, Dhaka. Presently, he is undergoing a
research programme on “Smart Intrusion Detection
System Comprised of Machine Learning and Deep
Learning” as part of Masters Programme in Information Security System, in
Bangladesh University of Professionals (BUP). He is also one of the authors
of an international publication, published in IEEE Digital Explorer titled
“Design and integrate dual renewable energy in a residential building of
urban area: A step towards the self-sustained smart energy system for
Bangladesh”DOI: 10.1109/ICEEICT.2014.6919139.

Asif Iqbal Khan earned his Bachelor of Science in

Engineering degree Information and Communication
Technology from Mawlana Bhashani Science and
Technology University in 2019. His major was
Information and Communication Technology.
Currently he is pursuing his Master’s degree in the
same institution. His research interests are in the field
of Artificial Intelligence, Machine Learning, Data
Science, Natural Language Processing, Predictive
analysis etc. He specializes in data analysis with tools like python, R etc.

Sajjad Waheed earned his doctorate degree on

Computer Engineering from the Istanbul University
in 2013. He is currently working as a Professor in the
Department of Information and Communication
Technology in the Mawlana Bhashani Science and
Technology University, Bangladesh. His research
interests are in the field of artificial systems, machine
learning, cryptography, data analysis, system
development, etc. He has received funds for his
research projects from the government and the university. He has published
more than twenty research papers in internationally acclaimed peer-reviewed
journals related to his research works.

DOI: http://dx.doi.org/10.24018/ejers.2020.5.10.2128 Vol 5 | Issue 10 | October 2020 6