HPSGNN: A Hybrid of Particle Swarm and Genetic Neural Network
System to Defense against Blackhole Attack Targeting MANETs
Tuka Kareem Jebur
Al-Mustansiriyah University, College of Management and Economic, Baghdad, Iraq
Keywords: Particle Swarm Optimization, Blackhole Attack, Mobile Ad-Hoc Networks (MANETs), Clustering.
Abstract: In this paper, propose a Hybrid of Particle Swarm and Genetic Neural Network system to the Defense
Against Blackhole attack Targeting MANETs. Detection and Prevention System black hole attack in
MANETs for this purpose two-stage applying the first stage using PSO to find an optimal cluster head this
reduces power consumption, conjunction, the second stage using the genetic algorithm to find optimal path
then used a neural network to detect and prevent malicious node in MANETs network with some criteria
nodes. Therefore, by isolated all data forms from the network, the Blackhole node is eliminated.
1 INTRODUCTION
When a router is affected by different causes. The
packet drop attack is very difficult to detect and
avoid because packets are regularly dropped from
the losses network (Vimal Kumar & Kumar, 2015).
MANET considers significant kinds in wireless ad
hoc network used in many fields but has a challenge
such as dynamic topology, no preexisting
infrastructure, packet drop, power consume
introducible change of device (node) can be moved
in any direction another drawbacks security issues
(Chaubey, 2015).
There is numerous protection to secure the
network from black hole attacks, such as a firewall,
to prevent unsafe activities. Such protections do not
however guarantee complete network security. The
second line of defense is therefore required that will
be capable of detecting new vulnerabilities every
day (K. Dalasaniya & N. Dutta, 2014).
Several systems have already been built for this
purpose, first used method splits the entire network
into interconnected structures called clustering the
continuous amount of nodes known as the (cluster)
CH responsible to aggregate data from the node and
send it to another CH or distention node so it
reduces cognition and power consume. Hence, when
using clustering in MANET reduces power
consumption and conjugation and another advantage
of the clustering method (N. J. Patel, 2015).
Unique routing protocols are used for this
purpose that can specify the route between nodes not
within each other's transmission range. particle
swarm optimization (PSO) consider one of the types
of computer algorithms based on the principle of
finding the value of the best solution among the
possible solutions to the problem depending on the
principle of experimentation and repetition. This
algorithm originated on the principle of the presence
of a swarm of elements. This squadron is spread in a
limited research area in the (problem space) and
moves randomly in the region to discover the best of
all solution in this region. The larger the number of
squadron elements and the smaller the search area,
the easier it is to find the perfect solution faster and
vice versa (A. Omidvar & K. Mohammadi, 2015)
genetic algorithm used in many research to
prevention Blackhole the attack, it can be defined as
one of the types of research methods can be
classified as an evolutionary algorithm .This
algorithm uses Darwin's revolutionary technology,
which includes inheritance, mutation, cross
crossover, and the production of the best solutions
by repeating the genetic cycle that is progressively
improved after each crossover, As a result of these
characteristics, are to find the best solution between
several solutions and the production of new
solutions have been used to solve some of the
problems facing the process of transfer or
confidentiality of data in wireless networks. (K.
Nikhil, S. Agarwal, and P. Sharma, 2012).
The research paper deliberates the concept of
detecting black hole attack and studied some of the
major detection methods such as the type of attacks
addressed and architecture in MANET. Furthermore,
artificial intelligence methodology called Neural
network considers an important method that can be
used to detect this attack paradigm attempts towards
mimic biological neural network structure and
functionality. The neuron takes the basic
construction block into account; the model includes
three easy rules: summing, multiplying, and
activation (A. Krenker, J. Bešter, & A. Kos, 2011).
In this type of attack, there are usually two cases
in which the data packet is obtained using a
malicious node; one where a malicious node uses the
routing protocol such as the AODV protocol, to send
route reaction control message (RREP) immediately
to the source node upon receipt of the route request
control node (RREQ). This RREQ overflow causes
unnecessary overhead that leads to reduced network
performance such as the delivery of packets and
latency. The success of the RREQ broadcast will
suffer from a hidden node problem. So it drops the
data directly. Therefore, the source node became un
incapable to send its data to the destination node
which disturbs the influence of the network and its
connectivity (L. Prajapati & A. S. Tomar, 2015). A
method to detect and prevention black hole attacks
in MANETs is proposed in this paper.
Moreover, by monitoring their neighbor's
actions, the system detects malicious nodes. If a
suspicious behavioral anode is found, declare the
suspicious node and send a threatening message. By
refusing all data forms, the black hole nodes are
isolated from the network. In the Network
Simulator, MATLAB simulations are performed to
test the performance of the technology proposed.
The results show that all types of black hole nodes
are identified and isolated by a proposed mechanism.
In this paper, focus on security challenges when
designing security schemes for MANETs, hence in
this work proposed a hybrid of PSO and GNN to
defense against Blackhole attack that targeting the
MANTs. The rest of the paper is organized in
section 2 are provided the relevant research. Section
3 outlines the scheme we are proposing while being
discussed in Section 3. 4, results, and desiccations
are shown. Last but not least, Section.5 is the
conclusion paper.
2
THEORETICAL AND
PREVIOUS STUDIES
Generally, the main assumption considered in the
MANET is that each node is a trusted node.
However, in a real scenario, some unreliable nodes
misbehave and launch the attack in a network like
Blackhole in which the misbehaving nodes attract all
the traffic towards itself by giving false information
of having the shortest path towards the destination
with a very high destination sequence number. This
section discussed different methods to detect or
prevent Blackhole attacks.
2.1
MANETs
MANETs have many users in many fields as such as
the modern technology revolution and its great
development, has a dynamic topology, no need for
infrastructure.
On MANETS, PSO and GNN algorithms were
suggested, each offering an effective implementation
technique. However, several investigators have
suggested various approaches to the black hole
attacks in MANETs. Most of these methods can be
classified into various categories such as the
following: In the work of (Omidvar & Mohammadi,
2014) the PSO algorithm has been suggested which
use the maximum flow objective to decide best node
locations for each network operation step, this
method adds some delay in the process time
discovery as intermediate nodes, computation time.
Presented technique by (Prajapati & Tomar, 2015)
this technique is called PSO of the AODV protocol
to find a solution for many network attacker nodes.
PSO tracks nodes by changing ad hoc values, if the
node converges then it switches node value to
endless and prevents the node from sending a
packet. This method has a drawback such as a delay
packed drop. However, the scheme needs to be
further analyzed, since values are modified after a
specific time period. Therefore, shorter update time
requires more overhead processing if accuracy of
detection decreases.
In the previous work of (A. Kaur, P. Kaur, & H.
Aggarwal, 2017) suggested using GA and PSO for
AODV routing protocols to detect the Blackhole in
WSN. By using this approach, it reduces power
consumption and finds the best bath from the source
to the destination suggested used genetic algorithm
(R. Garg & V. Mongia, 2018) with one type of
routing protocol called AODV for preventing the
Blackhole attack these methods need time toward
finding intruder node equal 13.2 whether suggestion
algorithm takes 0.64 times with 200 node The
scheme needs to be further analyzed, since values
are modified after a specific time period. So the
shorter update period requires more overhead,
otherwise the accuracy of detection will be reduced.
The used technique called clustering by Sanjeev
et al. (S. Gangwar, K. Kumar & M. Mittal, 2013) to
portion network to the region and using AODV to
opt node called a cluster head this method reduce
power consume in MANETs. (A. Augustine & M.
James,2015) present a method to extending and
butter performing network lifetime they suggested a
method based on ANN to detect the attack and using
AODV protocol to find path this method does not
prevent attack There is still the question of delay in
path exploration. In the work of (V. Kumar,2018)
presented a technique based Neural networks to find
malicious nodes in a Blackhole where a group of
nodes is examined according to the amount of
energy consumption, and the results are stored in a
table and updated periodically The results showed
that it's difficult to apply this approach in large
networks where nodes rapidly change positions.
The approach used in (F. Tseng, et al,2018)
present a method called GA was used to find
intruder nodes in a black hole attack A set of data
has been trained to identify this type of attack and
not to prevent the attack in minimum time.
As well (Kaur et al., 2017) proposed a safe route
discovery mechanism called GA to build the IDS
for black hole attacks in MANET. (Patel,
2015)The author uses GA for intrusion detection.
The GA-based IDS proposed analyses the behavior
of a node and identify black hole nodes based on
network parameters, e.g. packet drops, transmission
rate requests, and receipt rate requests, GA requires
time for evolution that is not suitable to detect
malignant nodes in MANET.
The present method by (Omidvar &
Mohammadi, 2014) it's focused on a forecast of
links and node lifetime algorithms in MANET. Path
recovery with PSO, this approach adds extra traffic
and allows more messages to prolong the discovery
process.
All of these and other works are aimed at finding
a route or finding a blackhole node for attack or
prevention. However, improved implementation in
MANETs.
3
THE RESEARCH METHODS
This section introduces the research methods which
consists of the system design, testing dataset, and the
evaluation methods, and it is described as the
following:
3.1
Testing Dataset
Several types of datasets have been used to evaluate
the performance of the Intrusion Detection System
(IDS) and Intrusion Prevention System (IPS) such as
KDD, CAIDA, BDD, and DARPA (Patel, 2015).
Among these different types of datasets, the
Berkeley Deep Drive (BDD) dataset is selected due
to its variety of features that could be suitable to
evaluate the performance of our proposed model.
The available BDD dataset in (Yassein, Khamayseh,
& Abujazoh, 2016) was utilized to evaluate the
HPSGNN system in this study. It is involving
feature selection to select the main as well as the
most appropriate features for detecting and
preventing Blackhole attack, the selection and the
removal of a redundant, non-relevant feature from
the data, to achieve an efficient and effective
selection process. A big behavioral feature of the
black-hole node that it presents itself as an
intermediate node with the best route towards the
target node, it sends an RREP message to the
destination node with a high destination sequence
number and a low number of hops. Thus, to
determine the "High Report Sequence Number,"
"Number of low hop number to destination"
characteristics, the thresholds for destination and
hop count are determined before the preliminary
characteristics are obtained. These thresholds are
determined by collecting the RREP data that each
network node transmits and recalculates, Moreover,
features are selected as relevant features for the
Blackhole attack.
3.2
The HPSGNN System Design
In this section, the system design has been
described. There are many types of attacks targeting
MANETs such as Routing Attacks, Black Hole
Attacks, Grey Hole attacks, Rushing Attacks (Tseng
et al., 2018). Furthermore, a blackhole Attack is the
most common type of attack targeting MANETs
(Gurung & Chauhan, 2019). In this work, we
attempt to overcome the Black Hole Attack and
protect the MANETs by proposing the HPSGNN
system. The HPSGNN system is implemented by
using MATLAB software, and 64 bit Windows 8.
The used computer has specifications of Intel CPU
core i7 @ 2.10 GHz with RAM of 4 GB. The
performance of the system is computed by analyzing
the results of the tested NAME Dataset. However,
the HPSGNN is consisting of a hybrid of (a) Particle
Swarm Optimization (PSO) algorithm, (b) Genetic
Algorithm, and (c) Neural Network as the following:
3.3
Particle Swarm Optimization
(PSO) Algorithm
It is a technology of evolution using a population of
candidate solutions to create an optimal solution to
the problem. A fitness function is used to calculate
the degree of optimality. It is motivated by collective
actions in societies with organized communities and
evolving intelligence. It uses several nodes
(particles) that make a swarm in the search area in
search of the best solution. This technique is used
here for the global optimization of the node values.
By converging the values using the shortest route of
the network nodes, PSO optimizes the values
(Preetha & Chitra, 2017). This article suggests a
solution by using the PSO-algorithm to optimize ad
hoc network numbers of clusters and energy
dissipation in nodes to provide an energy-saving
solution and minimize network traffic. This method
searching for a more effective and reliable solution.
Inter-cluster and intra- cluster traffic is handled by
cluster heads in the proposed solution. The
algorithm proposed takes into account node volume,
transmission power, and the mobile node battery
power consumption. This approach provides a
variety of options at a time (Kaur et al., 2017). The
main benefit of this algorithm provides a solution
with appropriate clusters and this method takes into
account various parameters such as ideal degree,
mobility, transmission power, and node capacity.
The CH is chosen on this basis and this CH is
responsible for interacting with the cluster nodes and
the neighboring CHs (Fahad et al., 2018). This
method has the main benefit of offering a series of
solutions simultaneously to the algorithm to find
optimal CH.
3.4
Genetic Algorithm (GA)
It is one type of machine learning, accompanied by
its operation as an example of nature's creation
cycle. In a population system, chromosomes show a
set of characteristics identical to base-4
chromosomes, which completes the creation. This
algorithm has 5 components.1. Population size: 2.
No. of variables: 3. Mutation 4. Crossover: how
much mutation happens is described. 5. Fitness: this
role ultimately decides the condom (Kukreja,
Dhurandher, & Reddy, 2018). To find the optimal
path GA takes the variety of PSO nodes from the
previous phase in a population recognized as
chromosomes. Growing chromosome is shown as 0s
or 1s bits. The selection is determined on each route
according to the maximum fitness value. If the
whole route has the highest fitness value, new
chromosomes will be picked using crossover from
this point. Crossover is also known as recombination
of two route path and finds a new path. Then here
also fitness is estimated and fewer Hops to hop
count distance nodes path is chosen as first. Here if
the node doesn’t satisfy the blackhole attack then
those nodes are well-thought-out as normal,
Mutation adjusts new chromosomes by changing
two bits in the node's position. A chromosome
picked for the mutation will have an arbitrarily
picked bit different from 0 to 1, or vice versa
(Dalasaniya & Dutta, 2014).
3.5
Neural Network (NN)
ANN is used for the detection of the node in the
suggested work. MATLAB's neural network is
initially trained according to the characteristics of
the network's nodes. ANN consists primarily of 3
layers, input, output, and hidden layer. The weights
are modified in the hidden layer to increasing the
variance between the input and the produced output
and therefore the desired output (Krenker et al.,
n.d.)The function activation is used for weighted and
input based outputs. If the result is consistent with
the actual output, the input is right therefore the
output would be modified to weight. The output is
compared to the target; when the route is discovered
between the source and destination node then the
attacker or intruders in the set route using the ANN
are being checked and if the attacker is being found
then their identification is saved. On the behalf of
the attacker‟s activities, the types of attackers have
been checked and the presentation from the attacker
is being checked to achieve better results (Preetha &
Chitra, 2017).
4
EVALUATION METHODS
In this analytical study, the performance of our
system is evaluated using eminent metrics, such as
packet delivery ratio (PDR), Detection Rate (DR),
and Throughput. PDR: calculate how much data can
be succeeding is reached to the recipient if send it.
DR represents the total number of detected nodes
(whether these are black hole nodes or not) from the
overall network traffics. Whereas, the Throughput
represents the process of calculating the number of
delivered data in seconds. Where in the PDR was
calculated as by Eq. 1:
(1)
Where in the DR was calculated as by Eq. 2:
(2)
Where in the Throughput was calculated as by Eq. 3:
(3)
One-way delay: processing of calculating the time to
send data from the sender to the recipient over the
network
One way delay = NL/R + ( P-1) L/R = (N+P-1)L/R
(4) N = link ,L = packet length ,R = transmission
rate
5
SIMULATION AND RESULTS
The suggestion proposed methods Applying PSO
and Genetic Artificial Neural Resource Networks
(G-ANN) for the defense of Blackhole attacks is
discussed in this section. Step I. Build a network
simulator environment with some basic dimensions
and important data set called DBB In the first case,
N nodes are generated within the MANET for
simulation to deploy the ad hoc mobile network.
Step 2. The source and destination nodes from the N
nodes have been described with their location after
simulator creation. Step III. Use the PSO algorithm
as the Cluster algorithm to find optimum no. of CH.
Every node scope, including source and destination,
was then initialized. Step IV. A code is generated to
define the path between source and target node for
the GA routing protocol. Step V-The GA algorithm
for the discovery of the route and the best selection
of the route by the scope is initialized. Step VI. The
fitness function of the GA algorithm is calculated
according to the information requested. Step VII.
When a route is discovered between source and
destination node then the attacker or intruders in the
set route using the ANN is being checked and if the
attacker is being found then their identification in
the routing table is saved. Step VIII. The types of
attackers have been tested and the presence of the
attacker is tested to achieve better results for the
operation of the attacker.
The GA and NN also reduce in comparison with
an attack the effect of the black-hole assault within
the. The parameter count used is energy use, packet
delivery ratio by changing the value of nodes, pause
time, location. Genetic algorithms have been
successfully applied for black hole avoidance and
optimization.
Figure 1: Simulation Model.
1 First phase finding the optimal number of CH
in MANETs network by using PSO where this
method portion network to the region and then
choosing CH node responsible to aggregate
data then forward it to destination.
2 The second phase takes the output from the
first phase and using Genetic find optimal path
and neural network algorithm to detect and
prevent attack with the number of parameters
that have been taken place in the network
using such as Throughput, Delay and the
dropped packets, delivery of the packet.
This section explains the findings achieved after
the proposed study was simulated. In the presence of
malicious nodes, the output of the proposed method
(AODV) is evaluated with GA. the method proposed
compared ADOV to our simulation findings and the
HPSO- GA (
Kukreja, D, 2018, Thanuja, R, 2018). The
main reason to choose these methods is that they are
the most recent strategies in the scientific literature
used and are close to our methodology,
i.e. the set of
nodes. The results also rely on measures of
performance such as detection rate of, packets drop,
throughput, and average delay. For every output
metric individually, the outcomes analysis is further
discussed.
Wireless network
node
PSO processing stage
GA processing stage
Node is normal
continua normal
procedure
Check the reason of
packed drop
(power consume,
link filer, congestion)
Blackhole node
start
No
Check node is
dropping
packed
Yes
End
Yes
If the node is
dropping full
packed
_
•
Detection Rate
Table 1 shows the results of the detection rate are
described in the approach suggested. The
identification rate is an important item for the
precise evaluation of malicious nodes in the status
packet. The reason to choose this measure is that the
proposed approach demonstrates the capacity of the
network to recognize the malicious nodes. On the x-
axis in figure 3, the number of nodes is shown, and
the y-axis indicates the detection rates (research
accuracy) of the AODV, GA, and HPSO-GA. This
indicates that the HPSO- GNN detection rate is the
highest (98%). This is because of the reaction from
every legitimate node to our proposed technology,
while the malicious nodes did not respond correctly
or drop down the status packet. The HPSO-GNN
strategy proposed would become more likely to
recognize malicious nodes as soon as the number of
malicious nodes increases in the network, the
proposed technology thus identifies malicious nodes
faster than other nodes and therefore improves the
detection rate with an increasing number of nodes,
which is the highest detection rate at 98.25%.
Table 1: The detection rate evaluation values with 200
nodes.
NUMBER OF
nodes
(HPSO-
GA)
(AODV) GA
Proposed
HPSO-GNN
10 93.21% 85.32% 87.97% 95.40%
50 93.65% 86.28% 88.18% 96.12%
100 93.91% 86.94% 88.29% 96.99%
150 94.58% 87.15% 88.63% 97.62%
200 95.13% 87.63% 89.35% 98.25%
Figure 2: Detection rate.
•
Packed Delivery Ratio
Table 2 and Figure 3 show the PDR of AODV, GA,
HPSO-GA, and HPSO-GNN.
The PDR of HPSO-GNN is the maximum
recorded (97.98%) because, after the detection of
malicious nodes, packets are simply delivered more
rapidly to the destination node. however, the
malicious nodes increase, certainly they will cover
the greatest of the network and will interrupt the
communication by sending fake replies and not
delivering data packets to the destination
appropriately. However, after the distribution of the
proposed technique, it was detected that the PDR is
considerably increased, as it blacklists the malicious
nodes with a status packet for a short amount of
time. Nevertheless, the PDR is much better when
compared to the other three protocols. Furthermore,
the result investigation of packet delivery shown
indicates that the proposed method outperforms
AODV, GA, and HPSO-GA.
Table 2: The packet delivery ratio evaluation values with
200 nodes.
NUMBER OF
nodes
(HPSO-
GA)
(AODV) GA
Proposed
HPSO-GNN
10 92.28% 81.12% 83.93% 95.56%
50 92.96% 82.85% 86.53% 96.52%
100 93.15% 84.16% 87.28% 97.32%
150 93.68% 84.63% 90.15% 97.42%
200 93.96% 86.24% 90.028% 97.98%
Figure 3: Packed delivery ratio.
Table 3: The one-way delay(s) with 200 nodes.
NUMBER OF
nodes
(HPSO-
GA)
(AODV) GA
Proposed
HPSO-
GNN
10 0.09 0.34 0.21 0.05
50 0.08 0.35 0.23 0.055
100 0.1 0.39 0.238 0.06
150 0.16 0.16 0.43 0.2 0.075
200 0.19 0.19 0.468
0.2
0.043
•
Throughput
Table 3 and Figure 4 showed the throughput of the
network when node is translation. However, in the
presence of malicious nodes in the network, as
shown in the figure the proposed technique and
AODV are compared. If there are two normal nodes
and any malicious nodes send the false routing
information claiming that there is a correct route
when the data packets actually are dropped, the
output is reduced. The mobility of the nodes also has
a direct impact on the network's efficiency, as the
movement of the nodes causes a breakdown in the
connection and results in a lower network output.
Table 4 demonstrates better results based on the
technology's efficiency (kbps). As in HPSO- GNN,
every node sends status packets to which each node
responds positively. When a node does not satisfy
the predefined criteria, then the node labels it as a
malicious node and the other nodes cease
communicating with it.
Table 4: The throughput evaluation values.
NUMBER OF
nodes
(HPSO-
GA)
(AODV) GA
Proposed
HPSO-GNN
10 69.16 59.23 66.288 73.122
50 70.21 60.18 66.97 74.44
100 71.76 62.42 67.71 75.91
150 72.63 63.12 68.55 76.61
200 72.34 63.78 68.75 77.33
Figure 4: Throughput with 200 node.
•
One-way Delay
Table 4 shows the one-way delay with a varying
number of nodes. Figure 5 shows the delay in the
results of the AODV and the proposed HPSO-GNN
using the nodes time to forward packets on time to
the intended node. The results show that the AODV
network delay is high because, during transmission,
the malicious node drops the data packets. In our
technique suggested, the delay at meager points is
slightly higher because the number of nodes
continuously sends packets to regular status nodes.
In MANETs, the connection failure was likewise
obvious, which implies that the nodes were to
retransmit the data packets back to the destination
node, so that time required could cause the delay It
can be seen that at meager points, in particular, The
HPSO- GNN proposed is more efficient than other
approaches.
Figure 5: One-way delay.
•
Packed Drop
Table 5 and Figure 6 show the results of a packet
drop in the network, with no. of nodes (200) the
packet drop rate proposed approach with puss time
(5-15) sec and No. of black hole nodes (1-3). in
which the AODV and the Proposed HPSO-GNN can
be compared. It can be seen in the proposed
technique that their Isa slight lag at a point, as the
number of nodes is decreased; the reason for this is
that the nodes end packets periodically. When a
packet has broadcasted the node, it takes less time to
reach every node in the set rather than the whole
network. Despite sending broadcast status packets
occasionally, the packed drop of the proposed
technique decreases. The overall performance of the
proposed method in terms of the packed drop is
decreased more than other methods.
Table 5: Comparative packet drop rate.
NUMBER OF
nodes
(HPSO-
GA)
(AODV) GA
Proposed
HPSO-GNN
10 1.10 3.12 4.11 0.01
50 2.11 3.98 4.5 0.10
100 3.12 4.55 4.89 0.19
150 4.13 5.32 5.28 0.28
200 5.14 5. 87 5.67 0.37
Figure 6: Packed drop rate with 200 node.
Table 6: Comparison of the proposed method with existing techniques.
Related work No. of CH Delay
Computation
Time
Throughput
Packed
Drop
Find Blackhole
node
Preeti and Sumita . x No
ShivaniI and Pooja . x x x
Kaural et al. x x x x
Garg and Mongia
. x
x
Gangwar et al.
x x
x
Augusti and James
. x x x
Kumar
. x
x
Tseng et al. x
x x
Fan-Hsun et al . x x
x
Shruti and Rakesh . x x x x
Proposed method
x x
x
6
CONCLUSIONS AND FUTURE
WORK
In this paper, the proposed method to detect and
prevent black hole attack in MANET, it is capable of
delivering packets to the destinations even in the
presence of malicious node while increasing
network size decreases the packet loss and increase
the security. End to end time taken to deliver the
packet take smaller than another approach (with 200
nodes). To make the result more accurate the
performance of these two and detection rate as
compared to GA, HPSO-GA, and AODV. Future
work will include increasing the number of
parameters such as accuracy and routing overhead
measured by the number of packets required for the
communication in the network. Used approaches
GA, HPSO-GA, and AODV the simulation result.
Besides, for cooperative black-hole attacks, the
proposed approach is equally successful. The
MATLAB 2016 simulation findings indicate that all
real black hole was found by the proposed method.
This improves network efficiency by reducing the
rate of decline, with low false-positive efficiency.
This method its better performance in almost all
parameters: throughput, end to end delay, packet
drop rate for performance measurement will provide
a more reliable and accurate result. Integrating the
genetic ANN with Fuzzy Logic can result in a more
efficient and faster Blackhole detection and
prevention mechanism.
ACKNOWLEDGMENTS
This work has not received any financial support
from any scientific institution.
REFERENCES
Augustine, A., & James, M. (2015).
ANN to Detect Network under Black Hole Attack, (Ideas),
15–18.
Chaubey, N. (2015). Performance Analysis of TSDRP and
AODV Routing Protocol under Black Hole Attacks in
MANETs by Varying Network Size, (i).
https://doi.org/10.1109/ACCT.2015.62
Dalasaniya, K., & Dutta, N. (2014). A Survey of Cluster
Management Techniques in MANETs, IV(December),
7–13. https://doi.org/10.1109/COMST.2005.1423333
Fahad, M., Aadil, F., Rehman, Z. ur, Khan, S., Shah, P.
A., Muhammad, K., Mehmood, I. (2018). Grey wolf
optimization based clustering algorithm for vehicular
ad-hoc networks. Computers and Electrical
Engineering, 70, 853–870. https://doi.org/10.1016/
j.compeleceng.2018.01.002
Gangwar, S., Kumar, K., & Mittal, M. (2015). Cluster
Head Selection in Mobile Ad-hoc Network (MANET)
Using ART1 Neural Network Cluster Head Selection
in Mobile Ad-hoc Network (MANET) Using ART1
Neural Network, (April).
Garg, R., & Mongia, V. (2018). Mitigation of Black Hole
Attack in Mobile Ad-Hoc Network Using Artificial
Intelligence Technique, 3(1), 1168–1174.
https://doi.org/10.1109/ ACCESS.2020.3004692
Gurung, S., & Chauhan, S. (2019). Performance analysis
of black-hole attack mitigation protocols under gray-
hole attacks in MANET. Wireless Networks, 25(3),
975– 988. https://doi.org/10.1007/s11276-017-1639-2
Kaur, A., Kaur, P., & Aggarwal, H. (2017).
Implementation of Blackhole attacks in WSN using
Genetic Algorithm and PSO, 10(4), 717–726.
Krenker, A., Bešter, J., & Kos, A. (n.d.). Introduction to
the Artificial Neural Networks.
Kukreja, D., Dhurandher, S. K., & Reddy, B. V. R. (2018).
Power aware malicious nodes detection for securing
MANETs against packet forwarding misbehavior
attack. Journal of Ambient Intelligence and
Humanized Computing, 9(4), 941–956.
https://doi.org/10.1007/s12652-017-0496-2
Kumar, Vimal, & Kumar, R. (2015). An Adaptive
Approach for Detection of Blackhole Attack in Mobile
Ad hoc. Procedia - Procedia Computer Science,
48(Iccc), 472– 479. https://doi.org/10.1016/
j.procs.2015.04.122
Kumar, Vinod. (2018). A Review on Detection of Black
hole Attack Techniques in MANET International
Journal of Advanced Research in A Review on
Detection of Blackhole Attack Techniques in
MANET, (April 2014).
Nikhil, K., Agarwal, S., & Sharma, P. (2012). Application
of Genetic Algorithm In Designing A Security Model
For Mobile, 181– 187. doi : 10.5121/csit.2012.2116
Omidvar, A., & Mohammadi, K. (2014). Particle swarm
optimization in intelligent routing of delay-tolerant
network routing, 1–8. https://doi.org/10.1186/1687-
1499-2014-147
Patel, N. J. (2015). Detecting pack et dropping nodes
using machine learning techniqques in Mobile ad-hoc
network : n, 468–472. doi: 10.1109/SPACES.2015.70
58308
Prajapati, L., & Tomar, A. S. (2015). Detection of Black
Hole Attack with Improved AODV Protocol in Manet,
3535–3540.
https://doi.org/10.15680/IJIRSET.2015.04050119
Preetha, V., & Chitra, K. (2017). Soft Computing
Techniques in Mobile Adhoc network : A Technical
Overview. International Journal of Current Research
and Review, 9(14), 23–25. https://doi.org/10.7324/
ijcrr.2017.9146
Tseng, F., Chiang, H., & Chao, H. (2018). Black Hole
along with Other Attacks in MANETs : A Survey,
14(1), 56–78. https://doi.org/10.3745/JIPS.03.0090
Yassein, M., Khamayseh, Y., & Abujazoh, M. (2016).
Feature selection for black hole attacks. Journal of
Universal Computer Science, 22(4), 521–536.
