International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249 – 8958, Volume-9 Issue-2, December, 2019

Detection and Prevention of Manet using Hybrid

SVM with Ann
Vishal Walia, Rahul Malhotra

 large-scale network is the latest development in

Abstract: Mobile Ad hoc Networks (MANET) have been telecommunications networks and technology [4,5]. The basic
exceptionally vulnerable against attacks because of the dynamic operation of routing mechanism is to perform two tasks: first,
and self-configurable nature of its system foundation. This kind of finding the optimal route, and second, sending data groups
wireless network is appropriate for temporary communication
linked due to its nature of less-foundation and there is no any known as packets over the network [6]. Also to optimize the
control of centralized manner. Design a routing mechanism that route Firefly algorithm is used and the selection has been
are security aware with higher QoS parameter is very competetive performed using ANN approach.
and the major tasks involved in ad hoc types of network as per the
limited power resources and their dynamic routing topology. This II. RELATED WORK
paper mainly focused on the design of a secure and trusts based
on-demand routing mechanism using Ad-hoc on demand distance The attack of kind black hole is a major issue and by
vector (AODV) protocol to compute trust-based produces path utilizing the trust-factor, the system becomes reliable.
initialed from source up to destination that will fulfill minimum For each and every node, a model based on trust is built inside
two end-to-end QoS parameters of network. So here, the the area of network. This presented system has implemented
generalized AODV routing protocol has been extended from
for IDS (Intrusion Detection system) to identify and mitigate
traditional routing mechanism to analyze the performance of this
model with combination of artificial intelligence concept. The MANET black-hole attacks. N. Arya et al. [7] prevented the
proposed ad hoc based routing mechanism is used to found network from attacks makes more secure. It facilitates the
possible routes that are prevented through trust adjacent position enhancement of the PDR and reduces overhead control by
of security validation protocols and enhanced link optimized route improving the performance of the routing protocol. The future
computes on the basis of Artificial Neural Network (ANN) as an lies in improving the table entries in the destination node for
artificial intelligence algorithm for well-organized more effective detection of wormhole nodes. here, network
communication in MANET. In addition, this research security by using effective techniques to prevent hybrid has
demonstrates the effectiveness of bio inspired Firefly Algorithm
also provided and improved DoS attacks by the help of a
(FFA) as an optimization approach with the consideration of
several performance QoS metrics of network. The results have novel algorithm. P. Gupta et al. [8] used the network in every
been measured in terms of throughput and PDR with SVM and field and the power of the clients assured about the growth of
ANN approach. It has been observed that the throughput and PDR MANET. MANET is defined like a distributed system that
measured using ANN approach is better compared to SVM do not requires a certain transmission network or backend
approach an average of 0.755 PDR value has been obtained using since it executes user connections with the available nodes. A
ANN approach. number of executions were carried out in MANET, ranging
from defense to multi-user gaming, which requires the
Keywords: Mobile Ad hoc Networks, d-hoc on demand distance network to be saved from various attacks and trespassers. The
vector, Artificial Neural Network, Firefly Algorithm, SVM, and
researchers used the ECBDS to store the network from the
PDR.
particular attack namely Byzantine attacks and the use of
I. INTRODUCTION resources. T. A. Kolade [9] haveimplemented a system
having mobile network operates taken as a node and is
termed as less-infrastructure and wireless network
MANET (Mobile Ad hoc Network) is defined as a set of having nodes are moves in free manner and positions can also
autonomous users who communicate in a relatively limited be varies . The resources utilization has been detected on the
bandwidth area. These networks are considered mobile nodes basis of a new algorithm for security level. The bandwidth
that are moving without base station [1]. Therefore, the utilization and storage for few node never depend up on the
communication process is control by these nodes itself. For threshold configuration being used properly. The attack
data transmission routing plays a very important role in named as Byzantine has been observed if the accused is not
MANET[2,3]. Routing is a mechanism for transferring data regarded in due time through some sources. A. Adnan et al.
directly from source to communication network. The main [10] have utilized various kinds of loads and simulation
reason is that previous networks are manageable and have an measurements. As a result, QoS metrics named as; PDR,
integrated environment; at the same time, a high-volume and Throughput and AE2ED are tested in the network being
deployed.
Revised Manuscript Received on December 15, 2019.
* Correspondence Author
Vishal Walia*, PHD Scholar IGKPTU, Jalandhar, India
.vishal.walia4p@[Link]
Dr. Rahul Malhotra, Director , Guru Teg Bahadur Khalsa Institute of
Engineering & Technology, Chhapianwali, Malout, India

Published By:
Retrieval Number: B3537129219/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B3537.129219 4463 & Sciences Publication
 Detection and Prevention of Manet using Hybrid SVM with Ann

The generated outcomes after the simulation showed that the III. PROPOSED WORK
performance of the network in black hole attacks has been
The model of proposed work has categorized among two
decreased predominantly in PDR due to the nodes discarding
sections: Route Discovery and its improvement, detection and
each of the data packets traversing the path. In fact,
throughput efficiency has significantly decreased in the life of mitigation of threats. Initially a network using n number of
malicious nodes. This can be happened due to the reason of nodes is deployed using coordinate geometry to position the
the lack of distribution to the recipient of the packets getting nodes within the network. The design process of network
transferred. If there is no any effect of black hole attack, there creation is described below.
is some amount of increment in the AE2ED. It has to
be acknowledged as suspicious nodes, before confirming
about the routing table, immediately transmitted the response.
In the QoS parameters some variations have been occurred
that can be showed the performance of network, mainly
decreased in the presence of a black hole attack. Jain et al.
[11] have explored a network using fuzzy based model. The
model has been utilized for the detection of black hole attack
through consuming AODV along with Fuzzy based approach.
The entire process has been performed using NS-2 simulator
by deploying 50 numbers of nodes that are moving in the area
about 1km. The network has been used to detect malicious
nodes up to 2 to 80 %, which is possible by employed fuzzy
logic as a classification approach. Chhabra et al. [12] have
implemented a approach based on fuzzy potential Threat
protocol to protect MANET from Black hole attack. The
designed protocol has helped to deliver data on time with
smaller packet drop. The main problem that has been find out
in this research is that the nodes taken decision as per the rank
list. This result in increase in storage modules included with
decreases the communication speed. Abdel-Azim et al. [13] Fig. 1 Create network
have presented a scheme to utilized Genetic algorithm as an
optimization approach with Fuzzy and Neural network to The designed ad hoc network with 12 numbers of nodes
minimize the influence of black hole attack in MANET. along with one source and one destination nodes are displayed
According to the obtained result it has been observed that the in Fig. 2. The source and the destination servers are
rate for packet delivery after getting the network secure from represented by the yellow and the pink color respectively.
black hole attack is less compared to the Intrusion Detection
System (IDS) used in the network. From the literature it has IV. MATH
been studied that, to prevent MANETs becomes a necessary
to their deployment and use as MANET is used in vast secure
areas where data security and interaction are essential.
Established wireless network methods have used to achieve
some degree of security. However, such kinds of options may
always be adequate; as the ad hoc network has its own security
issues due to these solutions can not be resolved. In such a
scenario, a protection solution should always be combined
with such an intrusion detection system to achieve an
appropriate degree of security. A comprehensive approach for
detecting intrusion included in mobile nodes of MANETs has
proposed. This system in based on behavioral-anomaly, that
makes it flexible, scalable, customizable, and reliable. ITuse
AN AODV routing protocol to verify our system by running
mobile node simulations. It is very difficult to build an
intrusion detection system for wireless ad hoc networks.
Wireless network's very existence, i.e. lack of fixed Fig. 2 Route Formation
infrastructure, makes it difficult for the network to obtain
After nodes deployment next step is to find route in between
audit information. The wireless network's limited resources
the source and, the destination server. The route is formed by
are critical parameters to be considered when developing the
applying AODV as a route creation mechanism. The process
Black Hole Attack Secure Framework. Many times make a
of finding route is discussed below.
distinction in between the false alarms and a true positive is
very difficult task. The SVM technique is taken to resolve this
issue.

Published By:
Retrieval Number: B3537129219/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B3537.129219 4464 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-9 Issue-2, December, 2019

A. Route Discovery
To discover a route, AODV is used as a routing protocol.
This is a reactive routing protocol, in which route is formed
only when the nodes required to transmit data. It works mainly
into two predefined phases: (i) Route request and (ii) Route
Discovery. Initially, Route request is sent by the transmitting
node throughout its coverage range in the MANET. When the
route request packet is received by the nearby node, it stores
the address in its route list table. If the address is not matched
with the nearby node, then it forwards the request packet to its
nearby node and an acknowledgement in terms of reply packet
send to the transmitter node that means current node is not a
destination. In this way, the packet reached at its destination
node through multiple possible routes.

Fig. 5 Prevented Network

SVM has considered to be that path along included its

related consumption of energy and the delay in the path as the
proposed design stored a records of every path and node.
Therefore it becomes a supervised learning process; the
classified route would be the similar as the practice
route transferred. In this particular case, if the element of the
path does not suit for the pointed element, it can be called a
suspicious route. The same design is used and listed for each
path. The only difference is that the power consumption for
the path is replaced by the node's power consumption and the
delay generated by the path is replaced by the delay created
through individual nodes.

B. USE OF ANN
In this research, three layer architecture of ANN is utilized
Fig. 3 Route Discovery Process using (AODV) for classification of the attack named as black hole by using
the concept of deep learning tool box of MATLAB. Following
is the input data, which is provided in the structure of input
layer of ANN. The input data has been forwarded to the
hidden layer in which the processing of data has been
performed. In case of unwanted results, the error, which is
termed as Mean Square Error (MSE) signal is feedback to the
hidden layer. Based on the input values, the network is trained
and later on used to detect black hole attack in the network.
The training structure obtained after the training process is
shown in Fig. 6.

Fig. 4 Affected Node

The attacker (black hole /gray hole) starts dropping data

packet and hence decreases the throughput. The affected node
is represented by gray color as depicted in Fig. 4. To
overcome this problem ANN along with SVM is used and the
route is diverse through the node, N11. In this way the
network is protected from the attacker. The proposed model
uses the power consumption of respondents and their network
operation time as the key source of light. It also determines the
light level by taking into account the lighting time. Any
respondent who meets the flying threshold shall be selected
and the less difference value should be selected as the
respondent.

Published By:
Retrieval Number: B3537129219/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B3537.129219 4465 & Sciences Publication
 Detection and Prevention of Manet using Hybrid SVM with Ann

17 Group (2) = Properties of training data according to

the non real nodes
18 else
19 Group (3) = Extra properties of training data
20 end
21 end
22 Initialization of ANN through Training data and then group
them
23 Net = Newff ( )
24 The parameters for training are set on the basis of the
requirements and perform training
25 Net = Train (Net, data for training and, Group)
Testing:
26 Current Node = Properties of the current node in the
network
27 Authorization = simulate (Net, Current node)
Fig. 6 Training Structure of ANN 28 if Authorization is valid
29 Authorized node = Right
Total number of nodes deployed inside the network is 50, 30 else
and the properties have been given as input data of ANN. 10 31 Authorized node = Wrong
number of neuron are passed to the hidden layer to adjust the 32 end
weights according to the requirement and output in terms of 33 Return; Authorized nodes as output
43 number of output data that represents the class of 34 end
communicating nodes.
The SVM and ANN is utilized in the detection of nodes of V. RESULTS AND ANALYSIS
black hole and gray hole in route using their optimized
properties by Firefly. Initially, SVM is utilized for the As per the structure of classification, some parameters are
detection of malicious node, after that validation has been analyzed.
performed using ANN approach. The algorithm of SVM with a) Throughput : This parameter is computed as total amount of
ANN is written below; Received Packets per unit time
b) True Detection Rate(TDR) : It defined as the total true
Algorithm: Support Vector Machine based Artificial detected intruders divided by Total amount of detections
Neural Network c) Packet Delivery Ratio(PDR) : It can be defined as the total
Input: Optimized Properties of nodes (T), Types of node number of packets that are received divided by total amount
(Cat), Kernel Function, and Neurons (N) of transferred packets
Output: Authorized Nodes The experimental result section is categorized into two
Training: sections (i) Results with SVM approach and (ii) Results with
1 Initialize the SVM training data T is the total nodes property ANN approach
with RBF as Kernel function
2 for I = 1 All nodes
3 if Property of Node (I) == Real
4 Defined the Cat as a category of training data
5 Cat (1) = Nodes (I)
6 else
7 Cat (2) = Nodes (I)
8 end
9 end
10 Train_Structure=SVMTRAIN (T, Cat, Kernel function)
11 T= Train_Structure. SupportVector //To find out the
training data for ANN
12 Initialize the basic parameters of ANN
– Number of Epochs (E) // Iterations used by ANN
– Number of neurons (N)
– Performance: Mutation, MSE, Gradient and
Validation
–Techniques to be utilized: Levenberg Marquardt
– Division of data: Randomly
13 for i = 1  T Fig. 7 Designed MANET Area
14 If T belongs to real nodes property
15 Group (1) = Properties of training data according to
the real nodes
16 else if T belongs to non real nodes property

Published By:
Retrieval Number: B3537129219/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B3537.129219 4466 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-9 Issue-2, December, 2019

packets/sec is attained by the system over the 20 simulation

2
Initially MANET area of 1000 m is designed by deploying rounds.
50 nodes in MATLAB as depicted in Fig. 7. It is observed that
at the present network is attacked by 4 attackers represented
by red triangles preventing the data transmission over the
nodes represented by green spheres.

Fig. 8 Average-PDR performance Fig. 10 Energy Consumption

The PDR determines the actual data packets delivered over Energy consumption by the system for 20 simulations is
the network. Fig. 8 depicts the average PDR performance of plotted in Fig. 10. The plot corresponds to multiple rise and
the MANET over 20 different simulation rounds. It is fall in the magnitude of energy consumption by the system
observed that initially the average PDR rises steeply to a value during simulations starting at 9mJ for 1st simulation round
corresponding to 0.654 for the first 3 simulation rounds. After and ending at 11mJ for the last simulation round. It is
this point, it rises from 0.65 to 0.675 with multiple observed that the system energy consumption ranges from
fluctuations for the next 4 to 20 simulation rounds with an 4mJ to 23mJ.
overall average PDR of 0.6862 attained for the 20
simulations.

Fig. 11 Residual Energy Changes

Fig. 9 Average-Throughput Variation
Fig. 11 depicts the changes in the residual energy
The average throughput variation of 20 simulations is corresponding to 14th simulation round over 50 nodes
shown in Fig. 9. The above Fig. corresponds to a steep rise of considered in the study. The plot shows that on average
540 packets per sec for the first 3 simulations. Later, residual energy of the system remains close to zero.
throughput of the network reaches a peak value of 640 packets
per sec at 6th simulation. Following this point, it gradually
decreases to 458 packets/sec till the end of 20th simulation
round. An overall average throughput of 453.8524

Published By:
Retrieval Number: B3537129219/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B3537.129219 4467 & Sciences Publication
 Detection and Prevention of Manet using Hybrid SVM with Ann

Comparative analysis of Average network throughput using reduction. The application of prevention algorithm
SVM and ANN is done based on three criteria, namely, re-establishes the network throughput to 59913 packets per
normal, under threat and after prevention. Table 1 summarizes time frame that corresponds to 65.32% improvement against
the average throughput obtained for the said cases for both the network under threat. It is observed that the throughput
SVM and ANN for simulation rounds ranging from 100 to obtained using ANN approach is higher as compared to
500. It is observed that average throughput obtained for ANN normal and under threat network.
is higher than SVM for all the three criteria.
PDR Evaluation
PDR evaluation is also done for normal network, network
under threat and network after prevention using SVM and
ANN. The resultant values of PDR under different conditions
are listed in Table 2 for 100 to 500 simulations.

Fig. 12 Average throughput improvement using SVM

Fig. 12 compares the average throughput of the network
under three circumstances namely the network acting normal
in the network, the network under a threat and the network
after the prevention using SVM. It has to be seen that the Fig. 14 Average PDR improvement using SVM
networks average throughput under the normal architecture is PDR evaluation under three conditions using SVM is
around 20875 data packets/ time frame. The measurable plotted in Fig. 14. It is observed that average PDR of the
metrics of this work named as throughput lowers down when normal structure is about 0.6603 which reduces to 0.3336
the system falls under threat. The throughput lowers down to when it falls under the threat. The improvements are attained
about 13825 packets per second which is a downgrade by due to the dual phase threat discovery process followed by the
33.77%. When the prevention algorithm is applied it goes route advancement architecture.
nearly normal to around 20775 packets per time frame
accounting to 33.45% improvement over the network under
threat. It is obvious that, if you prevent anything, you could
only attain what was the most normal, beyond that is not
attainable.

Fig. 15 Average PDR improvement using ANN

Fig. 15 represents the average PDR obtained for three
scenarios, namely, under normal, under threat and improved
after prevention using ANN as a classification approach. It is
observed that PDR for the designed network under normal
conditions is 0.6392 which get lowered by 45.61% under
threat condition corresponding to 0.3476. Application of
prevention algorithm increases the PDR to 0.7214 that shows
that ANN delivered data has higher packet rate as compared to
the packet rate obtained under threat condition.
Fig. 13 Average throughput improvement using ANN
Average throughput obtained with ANN under three
conditions is plotted in Fig. 13. It is observed that under
normal conditions average throughput of network is 41625
packets per time frame that get lowered to 20775 packets per
time frame under threat condition that corresponds to 50%

Published By:
Retrieval Number: B3537129219/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B3537.129219 4468 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-9 Issue-2, December, 2019

VI. CONCLUSION AUTHORS PROFILE

In the present work, routing mechanism for secure data Mr. Vishal Walia did his Bachelor as well as
transfer is presented for MANET that is susceptible to attacks. Masters in Electronics and Telecommunication
Engineering from IET, Bhaddal (Punjab Technical
Simulation study is conducted for 500 rounds to parameterize University, Jalandhar). He is currently a research
the energy consumption, residual energy, throughput and PDR scholar at IKG Punjab Technical University,
changes accompanying data transfer over the network. The Jalandhar.
network effectiveness is evaluated for data transfer using PDR His area of research are wireless and Mobile
Communication, Fuzzy Logic, Neural Network and
and throughput calculations under various conditions, mainly, Optimization. He has published more than 30
data transfer efficiency over network under normal conditions research papers in National and International Journal of repute. He is a life
is evaluated against the data transfer efficiency under attack member of ISTE.
and improved data transfer efficiency using SVM and ANN
classification approaches to reach a well organized Dr. Rahul Malhotra did his Bachelor of Electronics
and Telecommunication Engineering from Amravati
communication system. It is observed that throughput of the University Amravati, in year 2001. He did Masters of
normal network get improved by 33.45% and 65.32% with the Technology in Electronics and Communication
application prevention with SVM and ANN. The network Engineering from Giani Zail Singh College of
PDR also demonstrated an improvement of 45.33% and Engineering and Technology, Bathinda and
Doctorate of Philosophy in the faculty of
51.81% using threat prevention measures using SVM and Engineering and Technology from Punjab Technical
ANN. University in collaboration with Thapar University
Patiala.
REFERENCES His area of research includes Evolutionary Computing Techniques, Wireless
Communication systems. He started his professional career from HCL
1. Harjeet Kaur, VarshaSahni, [Link], “A Survey of Reactive, Technologies Bangalore and later he shifted to technical education industry,
Proactive and Hybrid Routing Protocols in MANET: A Review”, with elite educational groups of Punjab,
International Journal of Computer Science and Information He specializes in Wireless Adhoc Networks, Fuzzy Logic, Neural Network
Technologies, (IJCSIT), Vol. 4 (3), pp. 498-500, 2013. and Optimization. He has published more than 110 research papers in
2. Rao, R. L., Satyanarayana, B., & Kondaiah, “Performance of CBIDS National and International Journal of repute. He has guided more than 85
on AODV Routing Protocol against Black hole attacks in MANET”, research thesis at Master’s level and completed 04 Doctoral level of
IJSRCEIT, vol. 3, no. 3, pp. 1637-1644, 2018. Research. He is a Fellow Member of Institution of Engineers, Calcutta,
3. Panda, N., & Pattanayak B. K., “Energy aware detection and Institution of Electronics and Telecommunication Engineering, New Delhi
prevention of black hole attack in MANET”, International Journal of and senior member of CSI and ISTE.
Engineering and Technology (UAE), vol. 7 no. 26, pp. 135-140, 2018. He served as faculty Head of Electronics and Communication engineering
4. Pooja, V. S., Rohit, T., Reddy, N. M., & Sudeshna S., “Mobile Ad-hoc Guru Gobind Singh College of Engineering and Technology Talwandi Sabo,
Networks Security Aspects in Black Hole Attack”, in 2018 Second Adesh Institute of Engineering and Technology, Faridkot since 2001 to 2012
International Conference on Electronics, Communication and to teach Graduate and Postgraduate level of Engineering courses. In year
Aerospace Technology (ICECA) , pp. 26-30, 2018, IEEE. 2011, he started groundwork to establish a new venture of Adesh Group at
5. Farooq, M. U., Wang, X., Sajjad, M., & Qaisar S., “Development of Chandigarh, Adesh Institute of Technology, Chandigarh Campus, Gharuan
Protective Scheme against Collaborative Black Hole Attacks in Tricity, and in year 2012, he joined as Founder Director-Principal of the
Mobile Ad hoc Networks”, KSII Transactions on Internet and campus.
Information Systems (TIIS), vol. 12 no. 3, pp. 1330-1347.
6. Patel, M., Sharma, S., & Sharan, D. (2013, April). Detection and
prevention of flooding attack using SVM. In 2013 International
Conference on Communication Systems and Network
Technologies (pp. 533-537). IEEE.
7. N. Arya, U. Singh and S. Singh,”Detecting and avoiding of worm hole
attack and collaborative blackhole attack on MANET using trusted
AODV routing algorithm”,. In 2015 International Conference on
Computer, Communication and Control (IC4) IEEE, pp. 1-5, 2015.
8. P. Gupta, P. Goel, P. Varshney, and N. Tyagi, “Reliability Factor Based
AODV Protocol: Prevention of Black Hole Attack in MANET”,
In Smart Innovations in Communication and Computational Sciences,
Springer, Singapore, pp. 271-279, 2019.
9. T. A. Kolade, "A Scheme for detecting and mitigating cooperative
black hole attack in AODV-based MANET routing protocol." PhD
dissertation, 2018.
10. A. Adnan, A. B. Kamalrulniza, M. I. Channa, and A.W. Khan. "A
secure routing protocol with trust and energy awareness for wireless
sensor network", Mobile Networks and Applications, Vol. 21, No. 2.
pp 272-285, 2016.
11. Jain, A. K., Tokekar V. & Shrivastava S., “Security Enhancement in
MANETs Using Fuzzy-Based Trust Computation Against Black Hole
Attacks”, in Information and Communication Technology, pp. 39-47,
Springer, Singapore, 2018.
12. Chhabra, A., Vashishth, V., & Sharma, D. K., “A fuzzy logic and
game theory based adaptive approach for securing opportunistic
networks against black hole attacks”, International Journal of
Communication Systems, vol. 31, no. 4, pp. 3487-3510, 2018.
13. Abdel-Azim, M., Salah, H. E. D., & Eissa, “IDS Against Black-Hole
Attack for MANET”, IJ Network Security, vol. 20 no. 3, pp. 585-592,
2018

Published By:
Retrieval Number: B3537129219/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.B3537.129219 4469 & Sciences Publication