Content uploaded by Samuel Kofi Erskine
Author content
All content in this area was uploaded by Samuel Kofi Erskine on Apr 24, 2019
Content may be subject to copyright.
Content uploaded by Samuel Kofi Erskine
Author content
All content in this area was uploaded by Samuel Kofi Erskine on Apr 20, 2019
Content may be subject to copyright.
electronics
Article
Secure Intelligent Vehicular Network Using
Fog Computing
Samuel Kofi Erskine and Khaled M. Elleithy *
Computer Science and Engineering Department, University of Bridgeport, Bridgeport, CT 06604, USA;
serskine@my.bridgeport.edu
*Correspondence: elleithy@bridgeport.edu; Tel.: +1-203-576-4703
Received: 16 February 2019; Accepted: 18 April 2019; Published: 24 April 2019
Abstract:
VANET (vehicular ad hoc network) has a main objective to improve driver safety
and traffic efficiency. The intermittent exchange of real-time safety message delivery in VANET
has become an urgent concern due to DoS (denial of service) and smart and normal intrusions
(SNI) attacks. The intermittent communication of VANET generates huge amount of data which
requires typical storage and intelligence infrastructure. Fog computing (FC) plays an important
role in storage, computation, and communication needs. In this research, fog computing (FC)
integrates with hybrid optimization algorithms (OAs) including the Cuckoo search algorithm (CSA),
firefly algorithm (FA), firefly neural network, and the key distribution establishment (KDE) for
authenticating both the network level and the node level against all attacks for trustworthiness in
VANET. The proposed scheme is termed “Secure Intelligent Vehicular Network using fog computing”
(SIVNFC). A feedforward back propagation neural network (FFBP-NN), also termed the firefly neural,
is used as a classifier to distinguish between the attacking vehicles and genuine vehicles. The SIVNFC
scheme is compared with the Cuckoo, the FA, and the firefly neural network to evaluate the quality
of services (QoS) parameters such as jitter and throughput.
Keywords:
VANET; fog computing; DoS attacks and intrusion; FFBPNN; Cuckoo search algorithm;
firefly algorithm; QoS parameters
1. Introduction
It is noticeable that the automation industry has substantially improved in the last couple
of years. The integration of hardware and software components produces better drivability and
customer satisfaction. A vehicular ad hoc network (VANET) contains mobile vehicles with on-board
processing units (OBPU) and roadside units (RSUs) that assist vehicles [
1
–
3
]. Vehicle-to-vehicle
(V2V) communication is fortified to provide improved information to the drivers regarding roadside
accidents, traffic jams, etc. This improves driver safety and the driving comfort of the vehicle
in city traffic and on highways [
4
]. Highways, crossroads conditions, weather conditions, and
vehicles monitoring are now part of the VANET important safety applications that must be complied.
Examples of the safety applications include: Slow stop vehicle advisor (SSVA), post-crash notifications
(PCN), and collision/congestion avoidance (CCA). These safety applications are important for VANET.
VANET utilizes these safety applications to acquire prior knowledge of crossroads, highways, and
knowledge of other vehicles conditions. In addition, safety applications enable drivers to execute
sound judgment. Through safety applications, drivers are capable to obtain real-time information
needed in order to enable them to initiate logical judgment and prevent further road and highway
accidents occurrence. Regarding SSVA, vehicles that have slowed down or halted convey messages or
information while utilizing warning signals message received from the network and take appropriate
action. The warning signal messages sensitizes the surrounding vehicles in the VANET that may be in
Electronics 2019,8, 455; doi:10.3390/electronics8040455 www.mdpi.com/journal/electronics
Electronics 2019,8, 455 2 of 25
danger. With regard to PCN, messages are conveyed to highway patrols for further assistance through
neighboring vehicles. Neighboring vehicles are closer to each other such that trust establishment
in them becomes an urgent issue with VANET. The trust gained through neighboring nodes would
enable them to acquire accurate and real-time information of accidents and any emergency situation
on roads. It will also identify any denial of service (DoS) and intrusions on emergency activities
that may have been encountered in the network. Moreover, safety applications such as SSVA, PCN,
and CCA are connected with the RSU and are also deployed in VANET and connected at the traffic
management office (TMO). However, the connection of these safety applications with the RSU requires
improvement and efficient information delivery. The safety applications and the network devices can
function appropriately and also ensure timely notifications about any accidents and road emergency
situations. In addition, installation of VANET and appropriate deployment of the RSU with safety
applications can help disseminate and process warning messages accurately in real-time without delay.
Moreover, it is anticipated that warning messages can be conveyed in a timely manner through the
VANET, through which the message can be relayed to other vehicles. The warning messages are
usually generated at the TMO and may include notifications of DoS and intrusion attack activity in the
network. Some of the DoS and intrusion attacks include congestion/collision (CC), link breakdown, and
bad road conditions. CC of vehicles can occur in VANET at any time on the road due to the behavior of
the disabled vehicle or accident which requires immediate attention and notification on a timely basis.
In addition, some DoS intrusion attacks which this research investigates include smart and normal
intrusions (SNI) attacks. DoS and SNI attacks may cause link breakdowns in the network. DoS and
SNI attacks also overwhelm the network and block the entire V2V communication within VANET.
DoS and SNI attacks encountered in VANET become road threats. When these occur, they prohibit
VANET safety applications to function appropriately. In addition, they may lead to further attacks in
VANET, including the bad road conditions and highway congestion encountering of many vehicles.
This will also make it difficult for drivers to prevent road casualties in a timely manner. DoS, SNI, and
DRA (DoS resilience attacker) all have the tendency to overwhelm the RSU. DoS and SNI can also
exploit the RSU computational and communication resources and cause flooding with any requested
information. However, the intent of RSUs and their deployed safety applications is to be able to collect
and analyze the real-time information from vehicles. The information that is eventually received by
V2V communication should be appropriately analyzed and evenly distributed to other neighboring
vehicles, connected through VANET and safety applications on timely manner through the end-to-end
(E2E) communication process. The E2E communications process in VANET is important; however, E2E
communication may experience particular DoS and SNI attacks which can also overwhelm the RSU,
which would then require urgent attention. The RSU may waste computational time, especially when
it encounters false message or information. Therefore, the RSU requires an efficient and secure storage
method to safeguard it from being compromised when delivering vehicle to roadside unit (V2RSU)
and V2V messages in VANET [5].
In VANET, V2V and V2RSU communication storage solutions for propagating safety information
to nearby vehicles in a timely manner have been investigated using vehicular cloud and fog computing
(VCF) [
6
]. The VCF model has been developed to utilize VANET resources efficiently due to fog
computing (FC) and cloud-based logical interaction. Based upon VCF, grouped vehicles cooperate
and communicate with each other and dynamically share sensing, computation, and resources for
decision-making on the road, as well as for improving traffic management and road safety. There are
some examples of VCF applications that can be relied upon which include:
•Collecting local and highways traffic conditions from neighboring vehicles for planning routes.
•
Processing the big data traffic information through local and highway traffic management
authorities.
•
Critical collaborative events including road congestion, accidents, and all forms of attacks
(including DoS and SNI attacks) can be reconstructed.
Electronics 2019,8, 455 3 of 25
Although these application scenarios have utilized FC and cloud-based applications for efficient
storage and computations, this scheme has not been not appropriately secured. The authors claim that
their proposed scheme has achieved their aim in investigating quality of service (QoS) parameters in
VANET. Arguably, due to undetected DoS and SNI attacks, further investigation is needed. We believe
fog computing (FC) integration and the hybrid deployment of optimization algorithms (OAs) including
Cuckoo search algorithms (CSA), firefly algorithms (FA), firefly neural networks, and key distribution
establishment (KDE)/authentication sharing mechanisms is a promising solution for investigating
real-time data transmission and QoS parameters in VANET that answers to this question very well.
Thus, we believe integration of the KDE/authentication mechanism investigation for the network
level and the node level security can be achieved appropriately in order to ensure trustworthiness
of nodes and trustworthiness for the entire VANET. In addition, since RSUs play a major role in
distributing information in VANET, they can be secured appropriately to provide real-time end-to-end
V2V and V2RSU communication. Therefore, it has become urgent to investigate QoS parameters such
as delay/jitter and throughput in VANET. Moreover, due to the dynamic nature of VANET, it utilizes
a vulnerable wireless link. Wireless link deployment and connection with vehicles and associates
connect through multimedia safety applications should be secured when vehicles connect with the
RSU [
7
]. Since multimedia safety applications are now a part of the VANET system, however, they
are easily plagued by DoS and SNI attacks through the RSU. Multimedia safety framework demands
high QoS support and evaluation. QoS provision, in general, is required to supports the Media Access
Control (MAC) architectures [
8
]. MAC architectures for VANET rely on the VANET wireless medium
which can be implemented on DSRC (dedicated short range communication) data link technology [
9
].
In the past, researchers/authors have conducted several investigations on VANET. The authors’
investigation centered on multimedia safety application framework for determining QoS provision in
VANET, which also utilized FC for achieving the network level security protection, using the DSRC
data link technology. In addition, the authors have conducted separate investigations on OAs based
upon FC while utilizing DSRC data link technology for data transmission. The authors’ investigation
involved CSA [
10
–
12
], FAs [
13
–
15
] and a firefly neural network [
16
]. The aim of the authors was to
evaluate QoS parameters for delay/jitter and throughput in VANET. In addition, during the research
investigation, a firefly neural network was used to train effective misbehavior of the path delayed
in the VANET. Though the authors claimed to have succeeded investigating QoS performance in
the network, the QoS evaluation was not complete due to the inability of the researchers/authors to
consider the node level security evaluation in VANET. In addition, the authors did not investigate KDE
sharing, including hybrid integration with OAs. Therefore, there was a limitation in the evaluation
of trustworthiness in VANET, and both real-time information delivery and QoS provision within
VANET remain a major concern. FC integration with OAs including KDE sharing can be useful for
implementing VANET safety applications, since these schemes have the capability to ensure efficient
storage, time sensitivity, trustworthiness, and intelligence in real-time information delivery agendas
and QoS in Intelligent Transportation systems. To address these concerns, in this paper, we propose a
“Secure Intelligent Vehicular Network using fog computing” (SIVNFC) scheme for FC integration and
hybrid OAs deployment including CSA, FA, firefly neural networks, and KDE/authentication to detect
the network level and node level security in VANET against DoS, SNI, and other forms of attacks.
The main contributions of this research are:
•
Fog computing (FC) is integrated with hybrid OAs deployment including: CSA, FA, firefly neural
networks, and KDE. FC is used to determine the rapidly stored vehicular information. In addition,
the integration and deployment of FC with hybrid OAs and KDE provides intelligence which
reduces the search space for real-time information. It also prevents increased communication
times. Fog computing is an extension of cloud computing that provides computation, storage
services, and network communication services between the end nodes. The determination of the
rapidly stored vehicular information process relies on the communication behavior of vehicles in
this paper [17].
Electronics 2019,8, 455 4 of 25
•Secure the VANET at the node level and the network level for trustworthiness.
•
Determine reduced jitter and improved throughput for the VANET for real-time data transmission.
•
Use of regression model to confirm the accuracy of jitter/delay in the proposed SIVNFC scheme as
a road safety application.
The organization of the rest of the paper is as follows. Section 2presents related work. Section 3
discusses the DoS attacks, intrusions, and preventive mechanisms for the proposed SIVNFC model.
Section 4presents extensive simulation results and analysis of the results which includes: The feed
forward-backward propagation neural network, regression model, and QoS provision for the VANET.
Section 5is the conclusion including the future work of this research.
2. Related Work
In this section, VANET-related work is divided into two subsections which include: Section 2.1:
Securing VANETs-Centralized Architecture and Section 2.2: Securing VANETs-Fog Centric Distributed
Architecture as follows.
2.1. Securing VANETs-Centralized Architecture
The architecture of VANETs and their operations are comprehensively analyzed in the literature [
18
].
The data sharing and key distribution mechanism during the data transfer were studied in [
19
].
Route discovery mechanisms were also developed and presented in the same scenario. We classify the
security scenario at two levels: The security at the node level and the security at the network level.
The node level security is applied when the selection of the node for the data transfer is involved, such
as trusted node selection and the application of location-aware services [20].
In [21], the authors proposed location information verification cum security using a transferable
belief model (TBM) for Geocast routing in VANET at the network level security. The proposed protocol
included two level of location information verification. In the first level, tile-based techniques were
used to verify location information correctness, whilst in level 2, collective information concerning the
announced location information for each vehicle was obtained using TBM with the help of neighbor
list information through all neighbor vehicles. The limitation of the proposed protocol is that it did
not recommend any method for the network level security in order to evaluate trustworthiness in
VANET. Rather, the proposed protocol only disputed traditional security methods and only proposed
location information verification that was transferable in VANET. In addition, no appropriate storage
solution was offered on a real-time data transmission scheme. The authors in [
22
] proposed a dynamic
congestion control scheme (DCCS) for safety applications in vehicular ad hoc networks to determine
only the network level security. The proposed scheme is a means whereby the reliable and timely
delivery of data in safety applications can be ensured for road users and drivers. The proposed DCCS
scheme objective also included the broadcasting of safety messages in order to ensure reliability and
timely delivery of messages to all network neighbors. However, the disadvantage of the proposed
scheme is that DCCS is without a fixed infrastructure. Moreover, there was no trustworthiness and
efficient storage mechanism for the evaluation of real-time information in the network.
In [
23
], the authors proposed a location error resilient geographical routing (LER-GR) protocol for
vehicular ad hoc networks to detect only the network level security. In the proposed LER-GR protocol,
a Rayleigh distribution-based error calculation technique was utilized for evaluating error in location of
neighbor vehicles. Based upon the LER-GR protocol, the least error location information was used for
determining next forwarding vehicles. However, due to the dynamic mobility of VANET, the proposed
protocol should have recommended an efficient storage solution and intelligence for data exchange
in location information that would also ensure the reliability of data transmission. In addition, there
was no trustworthiness evaluation to assess vulnerabilities in the network for secure transmission of
location data. In [
24
], the authors proposed an algorithm that achieved secured time stable Geocast
(S-TSG) for VANET in a vehicular traffic environment for only the network level security. The proposed
Electronics 2019,8, 455 5 of 25
protocol was intended to detect vulnerabilities including DoS attacks in VANET, due to a decentralized,
open dynamic, as well as a limited bandwidth and control of overhead information. However, in
the proposed protocol, there was no investigation conducted to evaluate either efficient storage or
an intelligent and secure method solution in VANET for real-time data transmission. The protocol
limitation also included an absence in optimize real-time vehicular traffic environment information
processing. In [
25
], the authors proposed a geometry-based localization for GPS outage in a vehicular
cyber physical system (VCPS) (GeoLV) for network level security protection only. The proposed
localization technique was a GPS assisted localization which has the tendency to reduce location
aware neighbor constraints in cooperative localization. In addition, the proposed GeoLV utilized
mathematical geometry for estimating vehicle location and focused on vehicular dynamics and the
trajectory of the road. Based upon the proposed scheme, static and dynamic relocations were performed
to reduce the impact of a GPS outage on location-based services. However, the limitation of the
proposed GeoLV technique was that it does not guarantee trustworthiness, and no FC method for
efficient storage solution in VANET geometry-based localization for GPS outage in VCPS model was
recommended or proposed in the scheme. It can be realized that the node level security detection was
a major issue with the proposed schemes.
2.2. Securing VANETs-Fog Centric Distributed Architecture
Security at the network level is defined as when the data has to travel from the source to the
destination. Secured routing, key distribution, and the encryption of data packets fall under the
network security method. Fog computing is used to store the network data and to reutilize it to
accelerate network performance. In [
26
], the authors introduced fog computing to extend cloud
computing in the context of the middle fog layer among cloud and mobile devices and produce various
benefits. The authors utilized a key sharing mechanism for secure transmissions. In [
27
], the authors
further discussed the usage of fog computing by using an event-based data gathering scheme.
When a data transfer is called in the network, a node is summoned to perform some activity, and
an event occurs. A route discovery process contains ‘n’ events, including attaching hops from the
source to the destination. The addition of a hop also requires the identification of trustworthy nodes,
which utilizes optimization algorithms (OAs) to perform a successful operation to help solve this type
of issue in computer science [28].
This research paper specifically utilized a hybrid of optimized Cuckoo search algorithms
(CSA) [
10
], firefly algorithms, [
15
] and firefly neural algorithms [
16
] to investigate DoS and SNI
attacks. The investigation also detected the node level and network level security and mitigated the
attacks for trustworthiness in VANET. In [
12
], the authors also conducted an investigation about the
cognitive behavior of VANET for high-speed mobility of VANET. In the investigation, it was discovered
that VANET also experienced frequent topology changes. In addition, it was discovered that VANET
incurred memory storage challenges for allocating spectrum resources. Hence, in [
12
], the authors
proposed the improved adaptive binary Cuckoo search algorithm to investigate DoS attacks in VANET.
The researchers in [
15
] used the firefly algorithm to investigate vehicles that travelled along highways
which encountered some form of VANET attacks. These vehicles that were deployed in the VANET
were vulnerable due to DoS attacks which caused delays at the network level. Afterward, the authors
utilized a clustering algorithm to facilitate good communication links. The authors’ investigation
centered on the real-time communication of the VANET to determine the efficiency of the messages for
vehicles in order to receive traffic warnings in a timely manner. The authors’ investigation conducted
on the FA was also used to determine the reliability of the warning signals. The authors also conducted
research in the FA and utilized the vehicles road-side infrastructure (RSU) regarding traffic safety
warnings. In [
16
], the authors utilized the firefly neural algorithm, which is a combination of FA
and a neural network, to investigate and train the VANET to determine the delay of the network.
The parameters used for training the VANET was used to detect the network level DoS attacks, and
the delay was evaluated in the network. The firefly neural algorithm utilized a machine learning
Electronics 2019,8, 455 6 of 25
process studied in VANET to determine the misbehavior of the vehicles/nodes for detecting DoS
attacks. The model consisted of four main phases including data acquisition, data sharing, analysis,
and decision making. Hybrid OAs deployment including CSA, FA, and the firefly neural network, can
integrate with fog computing and KDE to determine the node level and network level security against
DoS and SNI attacks.
Hybrid OAs deployments select the best solutions or minimize unnecessary solutions to retain
the contrast of the objective function. OAs are either heuristic or metaheuristic in nature. The heuristic
approach has problem-solving skills but is not suitable for each domain. NP-hard problems fall under
heuristic optimization algorithms. Non-heuristic algorithms are adaptive in nature and may be applied
for different sets of problems. More elaborately, the optimization can be further classified as Natural
Computing, Swarm Intelligence, or Medical Computing. Both the CSA, FA and firefly neural network
are classified as a Swarm Intelligence algorithms. There are various practices and architectures for
the CSA, FA, and firefly neural swarm intelligence (SI) algorithms that relate to the CSA used in this
research. In one practice or behavior, the Cuckoo bird lays its eggs in other birds’ nest and leaves its
eggs to be cared for by other bird species. In another behavior, a Cuckoo destroys all of its eggs, even if
only one egg is damaged, due to it considering that the eggs are not suitable for further reproduction.
In addition, this research paper has utilized the second behavior of the CSA in combination with
Lagrange’s method and the other swarm intelligence algorithms such as FA and firefly neural network
to select trustworthy nodes and ensure that the entire VANET is secure. The description is given in the
subsequent section. The network may suffer from different kinds of intrusions or attacks. One of the
most common security threats is the Denial of Service (DoS) and the SNI. In [
29
], different structures of
DoS attacks that also address the concern of SNI are discussed and presented. A detailed description
of DoS and the SNI attacks are also given in Section 3of this research paper.
3. DoS Attacks, Intrusions and Prevention Mechanism in VANET
3.1. DoS Attacks
VANET experience DoS attacks [
30
]. These attacks intercept the channel at the data link layer.
DoS attacks are capable of bringing down the available network resources. Through DoS attacks, the
VANET can be exploited through the RSU due to the following:
•
Resources consumption: DoS attacks consume the available network bandwidth. They inject fake
routing messages, resulting in congestion over the VANET. This degrades the end communicating
entities performance and introduces jitter.
•
Signal jamming: DoS attacks have a high tendency to jam the transmissions while using
channel interference.
•
Packet Drops: DoS attacks have a high tendency to drop all or any selected packets. This interrupts
the routing process from the source to the destination communicating entities.
•
The investigation of VANET security provisions, such as certificate-based identification and a
authentication mechanism are beyond the scope of this research.
3.2. Attack Principles
Unlike wired architectures where the channelblockage or congestion is always due to the increased
flow rates at links with bottlenecks, congestion in a VANET may occur due to the aggregation property
of the vehicles. If the attacker densely aggregates his attacks near the victim, the attacker can occupy
more communication channels [
31
]. The total transmission capacity of one node increases a linearly
with the increase in the area. If the node count does not vary, then the hop capacity is O (k), where k is
the node count of the network. The data transfer requires a route discovery, and the node count in
a route may increase with the increase in the area. Each node has a probability of 1/k of interacting
with the channel. There are m nodes that can act as attacking nodes such that the victim node has the
Electronics 2019,8, 455 7 of 25
likelihood of (1
−
m/k) of interacting with the channel. Figure 1illustrates the channel occupancy and
interaction of the proposed model architecture [32].
Electronics 2019, 8, x FOR PEER REVIEW 7 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
probability of 1/k of interacting with the channel. There are m nodes that can act as attacking nodes
such that the victim node has the likelihood of ( 1 − m/k) of interacting with the channel. Figure 1
illustrates the channel occupancy and interaction of the proposed model architecture [32].
6
5 1
4
Attacker
Vehicle
(node)
3
2
V2V DSRC
signal
Broken
signal
Figure 1. Channel occupancy.
VANET utilizes IEEE 802.11 as the most popular V2V DSRC (vehicle-to-vehicle dedicated short
range communication) wireless system installed on almost every vehicle where the vehicle/channel
congestion/collusion are inevitable due to influence of the attacker vehicle encounter in the network,
which could occur at the time when vehicles (or V2V) are required to transmit packets to each other
in VANET. CSMA/CA (Carrier Sense Multiple Access/Collision Avoidance) is standard scheme that
can be used to avoid such vehicle/channel packet transfer collision/congestion. However, CSMA/CA
is only a simple mechanism that can be used to allocate radio resources. In this research, we
investigated how vehicle/channel occupancy can cause delayed packet transmission due to
misbehavior of attacker vehicle which leads to broken link exposure of the vehicles communication
process as shown in Figure 1. Figure 1 illustrates channel occupancy scenario based upon the attacker
mode of operation.
In Figure 1, there are two types of vehicles, namely attacker vehicle and normal vehicle. All
attacker vehicles have broken signals connections with each other. When attacker vehicle forms a
connection with normal vehicle, a delay can be experienced in the network due to channel occupancy
as a result of broken signal connection because both normal vehicles and attacker vehicles are in each
other’s communication range and the vehicles are traveling on the highway. The first ellipse from left
to right has a transmission range (250 m), whereas the next ellipse has an interference range (550 m).
Attacker 2 transfers packets to vehicle node 3, and this processes are highlighted in a broken V2V
DSRC communicating signal, in which the packet is not received by another normal corresponding
vehicle. Now, vehicle nodes 5 and 4 are in the range of vehicle node 3, but since it is occupied by
attacker 2, it will have to wait, and an unnecessary delay will occur in the network. The channel
occupancy vehicular attacker scenario is also used to illustrate the misbehavior of compromised
nodes in VANET due to DoS attacks [33].
3.3. DoS Attack Illustration
A DoS attack employs multiple vehicles to attain its goal. It locks the job queue of the
corresponding vehicle so that it is unable to accept data packet requests from genuine vehicles. Since
a DoS attack is distributed over several vehicles, distinguishing authentic users becomes complicated.
There are several ways to mitigate the effects of this type of attack, including encryption and the use
of classification techniques [30]. The use of authentication mechanisms can also be beneficial. Sanya
Figure 1. Channel occupancy.
VANET utilizes IEEE 802.11 as the most popular V2V DSRC (vehicle-to-vehicle dedicated short
range communication) wireless system installed on almost every vehicle where the vehicle/channel
congestion/collusion are inevitable due to influence of the attacker vehicle encounter in the network,
which could occur at the time when vehicles (or V2V) are required to transmit packets to each other in
VANET. CSMA/CA (Carrier Sense Multiple Access/Collision Avoidance) is standard scheme that can
be used to avoid such vehicle/channel packet transfer collision/congestion. However, CSMA/CA is
only a simple mechanism that can be used to allocate radio resources. In this research, we investigated
how vehicle/channel occupancy can cause delayed packet transmission due to misbehavior of attacker
vehicle which leads to broken link exposure of the vehicles communication process as shown in
Figure 1. Figure 1illustrates channel occupancy scenario based upon the attacker mode of operation.
In Figure 1, there are two types of vehicles, namely attacker vehicle and normal vehicle. All attacker
vehicles have broken signals connections with each other. When attacker vehicle forms a connection
with normal vehicle, a delay can be experienced in the network due to channel occupancy as a result
of broken signal connection because both normal vehicles and attacker vehicles are in each other’s
communication range and the vehicles are traveling on the highway. The first ellipse from left to
right has a transmission range (250 m), whereas the next ellipse has an interference range (550 m).
Attacker 2 transfers packets to vehicle node 3, and this processes are highlighted in a broken V2V
DSRC communicating signal, in which the packet is not received by another normal corresponding
vehicle. Now, vehicle nodes 5 and 4 are in the range of vehicle node 3, but since it is occupied by
attacker 2, it will have to wait, and an unnecessary delay will occur in the network. The channel
occupancy vehicular attacker scenario is also used to illustrate the misbehavior of compromised nodes
in VANET due to DoS attacks [33].
3.3. DoS Attack Illustration
A DoS attack employs multiple vehicles to attain its goal. It locks the job queue of the corresponding
vehicle so that it is unable to accept data packet requests from genuine vehicles. Since a DoS attack is
distributed over several vehicles, distinguishing authentic users becomes complicated. There are several
ways to mitigate the effects of this type of attack, including encryption and the use of classification
techniques [
30
]. The use of authentication mechanisms can also be beneficial. Sanya Chaba et al.
(2017) presented a VANET architectural design for authentication key delivery with less delay between
vehicles and with more mobility by utilizing fog and cloud computing. The authors have also
Electronics 2019,8, 455 8 of 25
introduced fog computing to extend cloud computing to the context of the middle fog layer among
cloud and mobile devices for the production of various benefits. In their work, Qi Jian et al. (2018)
identified the security goals for VCC (vehicular cloud computing) interoperability.
The authors have provided the AKA (Authentication and Key Agreement) framework for VCC.
Notably, the authors have proposed the problems with the challenges for designing a consistent AKA
with extra strong security assurance for VCC. A hybrid AKA framework has been suggested that
combines the ‘single server 3-factor protocol’ with the ‘non-interactive identity-based key established
protocol,’ which computes the performance by a simulated platform. Fog computing is utilized quite
often these days for deployment of VANET, but its implementation has not been deployed with any
KDE or key sharing for preventing SNI attacks, also utilizing the RSU. Figure 2illustrates an attack
model scenario with the integration of the fog server with vehicles. In Figure 2RSU stands for road side
unit. The fog server keeps the information about the vehicles and distributes the required information
to other vehicles if required. The intruder may also utilize the same server and may misuse the server’s
information to spread false information [33].
Electronics 2019, 8, x FOR PEER REVIEW 8 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
Chaba et al. (2017) presented a VANET architectural design for authentication key delivery with less
delay between vehicles and with more mobility by utilizing fog and cloud computing. The authors
have also introduced fog computing to extend cloud computing to the context of the middle fog layer
among cloud and mobile devices for the production of various benefits. In their work, Qi Jian et al.
(2018) identified the security goals for VCC (vehicular cloud computing) interoperability.
The authors have provided the AKA (Authentication and Key Agreement) framework for VCC.
Notably, the authors have proposed the problems with the challenges for designing a consistent AKA
with extra strong security assurance for VCC. A hybrid AKA framework has been suggested that
combines the ‘single server 3-factor protocol’ with the ‘non-interactive identity-based key established
protocol,’ which computes the performance by a simulated platform. Fog computing is utilized quite
often these days for deployment of VANET, but its implementation has not been deployed with any
KDE or key sharing for preventing SNI attacks, also utilizing the RSU. Figure 2 illustrates an attack
model scenario with the integration of the fog server with vehicles. In Figure 2 RSU stands for road
side unit. The fog server keeps the information about the vehicles and distributes the required
information to other vehicles if required. The intruder may also utilize the same server and may
misuse the server’s information to spread false information [33].
Traffic Management
Office
Fog Ser ver
Lege nd
Normal vehicles
on route
Intr uder or co llided
vehicles
RSU
Vehicles utilize inter-
vehicl es communica tion
and DSRC technology
Fog ser ver
diss eminate inter
vehicles
inform ation and
DSRC technology
Figure 2. Intrusion/attacker model.
3.4. Intrusion /Attacks Model
Figure 2 illustrate the intrusion/attacker model (IAM). The model detects and mitigates Dos and
SNI attacks. The proposed IAM utilizes two types of vehicles, namely normal and intruder or collided
vehicles. Normal vehicles are supposed to be on route. Normal vehicles denote all vehicles that have
not experienced any form of attacks. Normal vehicles are the type of vehicles that are expected to
arrive at their destination safely. The intruder or collided vehicles, on the other hand, are the type of
vehicles that have encountered intrusion attacks. Normally they are not expected to arrive to their
destination. Moreover, the intruder vehicles have the tendency to introduce delays in the network.
If intruder or collided/disabled vehicles are left unattended and continue to remain in the
network, the network will suffer link breakdown and will not function as expected. This will lead to
much delay encounter in the network. Delays of the network will lead to further road casualties since
vehicles will not be appropriately informed. The proposed IAM initiates a remedy to prevent
Figure 2. Intrusion/attacker model.
3.4. Intrusion /Attacks Model
Figure 2illustrate the intrusion/attacker model (IAM). The model detects and mitigates Dos and
SNI attacks. The proposed IAM utilizes two types of vehicles, namely normal and intruder or collided
vehicles. Normal vehicles are supposed to be on route. Normal vehicles denote all vehicles that have
not experienced any form of attacks. Normal vehicles are the type of vehicles that are expected to
arrive at their destination safely. The intruder or collided vehicles, on the other hand, are the type of
vehicles that have encountered intrusion attacks. Normally they are not expected to arrive to their
destination. Moreover, the intruder vehicles have the tendency to introduce delays in the network.
If intruder or collided/disabled vehicles are left unattended and continue to remain in the network,
the network will suffer link breakdown and will not function as expected. This will lead to much delay
encounter in the network. Delays of the network will lead to further road casualties since vehicles will
not be appropriately informed. The proposed IAM initiates a remedy to prevent intruders/attackers
in order to lessen road casualties. Therefore, in the proposed IAM, vehicles utilize intervehicle
Electronics 2019,8, 455 9 of 25
communication and DSRC technology. The vehicles communicate and share safety information with
each other vehicle.
The information shared include condition of the vehicle and the road conditions. The information
shared may also include congestion/collision and accidents that have already occurred. In addition,
the fog server (FS) is deployed such that it addresses the location awareness concern in the cloud.
The deployed FS disseminates emergency inter-vehicles information utilizing warning sign to alert
other vehicles through the RSU information processing. The warning signal information can be
obtained by each vehicle through the RSU and the FS which originates from the traffic management
office (TMO). The TMO is the place where road safety applications (RSA) such including as SSVA, PCN,
and CCA are deployed and connected with the RSU and the FS. Two inter-vehicle communications,
including the FS, utilize DSRC technology. DSRC technology is data link technology which utilizes
the IEEE 802.11 standard for transmitting information. Based upon this, real-time information, which
convey warning and emergency information about any intruder activity in the network, can be received
through the RSA.
The network is also identified with the other forms of intruder/attacker such as smart and normal
intrusion (SNI). SNI may sometimes go unnoticed and requires sophisticated approach to detect.
Smart intrusions make the network feel like there is no threat in the network. If the intrusion follows a
set pattern of dumping the packets, then it becomes easy to identify. However, the smart intrusions do
not follow a consistent pattern [
34
]. The SNI scenarios that occur in the VANET are depicted as in the
figures below.
3.4.1. Smart and Normal Intrusion/Attacks Scenario
Figure 3a,b represent the normal and smart intrusions (SNI) attacker scenarios. The proposed IAM
relies on the SNI intensity to evaluate the delay of the network. The intensity and location of the normal
intrusion does not change with the change in the time frame, whereas the smart intrusion changes
the location and intensity of the attacks with every instance. As shown in Figure 3b, the Intrusion
is at location (x,y) at time t =0, and it instantly changes its position at time t =1 and goes to (x +t)
and (y +t). The intrusion even changes the location and intensity of the attack at every instance [
35
].
The SIVNFC system architecture prevention mechanism (SAPM) is a sophisticated approach that can
be utilized to determine and mitigate the SNI attacker in the VANET as demonstrated below.
Electronics 2019, 8, x FOR PEER REVIEW 9 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
intruders/attackers in order to lessen road casualties. Therefore, in the proposed IAM, vehicles utilize
intervehicle communication and DSRC technology. The vehicles communicate and share safety
information with each other vehicle.
The information shared include condition of the vehicle and the road conditions. The
information shared may also include congestion/collision and accidents that have already occurred.
In addition, the fog server (FS) is deployed such that it addresses the location awareness concern in
the cloud. The deployed FS disseminates emergency inter-vehicles information utilizing warning
sign to alert other vehicles through the RSU information processing. The warning signal information
can be obtained by each vehicle through the RSU and the FS which originates from the traffic
management office (TMO). The TMO is the place where road safety applications (RSA) such
including as SSVA, PCN, and CCA are deployed and connected with the RSU and the FS. Two inter-
vehicle communications, including the FS, utilize DSRC technology. DSRC technology is data link
technology which utilizes the IEEE 802.11 standard for transmitting information. Based upon this,
real-time information, which convey warning and emergency information about any intruder activity
in the network, can be received through the RSA.
The network is also identified with the other forms of intruder/attacker such as smart and
normal intrusion (SNI). SNI may sometimes go unnoticed and requires sophisticated approach to
detect. Smart intrusions make the network feel like there is no threat in the network. If the intrusion
follows a set pattern of dumping the packets, then it becomes easy to identify. However, the smart
intrusions do not follow a consistent pattern [34]. The SNI scenarios that occur in the VANET are
depicted as in the figures below.
3.4.1. Smart and Normal Intrusion/Attacks Scenario
Figures 3a and 3b represent the normal and smart intrusions (SNI) attacker scenarios. The
proposed IAM relies on the SNI intensity to evaluate the delay of the network. The intensity and
location of the normal intrusion does not change with the change in the time frame, whereas the
smart intrusion changes the location and intensity of the attacks with every instance. As shown in
Figure 3b, the Intrusion is at location (x,y) at time t=0, and it instantly changes its position at time t=1
and goes to (x+t) and (y+t). The intrusion even changes the location and intensity of the attack at every
instance [35]. The SIVNFC system architecture prevention mechanism (SAPM) is a sophisticated
approach that can be utilized to determine and mitigate the SNI attacker in the VANET as
demonstrated below.
(a) (b)
Figure 3. (a) Normal intrusion/attacker; (b) Smart intrusion/attacker.
Electronics 2019,8, 455 10 of 25
3.5. Proposed SIVNFC System Architecture Prevention Model (SAPM)
Figure 4depicts the proposed SAPM. In SAPM, vehicles utilize two DSRC technology instances
for information transmission (the DSRC technology uses the IEEE 802.11 standard for transmitting
information). In one instance of information transmission, vehicles communicate among themselves
using intervehicle or V2V communication. In the other instance, the FS forms a connection with the RSU
and, through this arrangement, disseminates inter-vehicle information to all vehicles in the network.
The information conveyed usually include collusion/congestion, intruder activity of the network such
as SNI of vehicles, or information of vehicles that have encountered attacks. The disseminated vehicle
information may also include reporting the state of vehicles conditions and the road conditions that
are threatening.
The proposed SAPM also employs further preventive measures to detect and mitigate all forms
of attacks, including DoS attacks that may go unnoticed. Some of these attacks include but are not
limited to packet drop, jamming of channels, and the RSU resources consumption overutilization.
Two models are deployed in the SAPM, namely IAM and VANET structure with integrated for server
(VSIF) models. The models utilize steps and scenarios for prevention and protection of the network
against DoS and SNI attacks. In step 1, collided/disabled vehicles or intruder activity are detected and
reported to the other vehicles in the network utilizing the IAM. The IAM detection of intruder/attacker
has already been explained in detail above. In scenario 2, the VSIF model is deployed. The deployment
of the VSIF model is also illustrated in Figure 5. The VSIF model relates and connect with the proposed
SAPM as below.
Electronics 2019, 8, x FOR PEER REVIEW 10 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
Figure 3. (a) Normal intrusion/attacker; (b) Smart intrusion/attacker.
3.5. Proposed SIVNFC System Architecture Prevention Model (SAPM)
Figure 4 depicts the proposed SAPM. In SAPM, vehicles utilize two DSRC technology instances
for information transmission (the DSRC technology uses the IEEE 802.11 standard for transmitting
information). In one instance of information transmission, vehicles communicate among themselves
using intervehicle or V2V communication. In the other instance, the FS forms a connection with the
RSU and, through this arrangement, disseminates inter-vehicle information to all vehicles in the
network. The information conveyed usually include collusion/congestion, intruder activity of the
network such as SNI of vehicles, or information of vehicles that have encountered attacks. The
disseminated vehicle information may also include reporting the state of vehicles conditions and the
road conditions that are threatening.
The proposed SAPM also employs further preventive measures to detect and mitigate all forms
of attacks, including DoS attacks that may go unnoticed. Some of these attacks include but are not
limited to packet drop, jamming of channels, and the RSU resources consumption overutilization.
Two models are deployed in the SAPM, namely IAM and VANET structure with integrated for server
(VSIF) models. The models utilize steps and scenarios for prevention and protection of the network
against DoS and SNI attacks. In step 1, collided/disabled vehicles or intruder activity are detected
and reported to the other vehicles in the network utilizing the IAM. The IAM detection of
intruder/attacker has already been explained in detail above. In scenario 2, the VSIF model is
deployed. The deployment of the VSIF model is also illustrated in Figure 5. The VSIF model relates
and connect with the proposed SAPM as below.
Step1: Collided
and nor mal
vehicles
VANET struct ure
with integrated fog
server
2
Step 2:Fog server collect and
distri bute disabled and
collided vehicles
information from the RSU
using inter-Vehicle
communica tion
1
Smart an d normal
Intrusion M odel
Fog Server
integrat ed
with cloud
City Traffic
Management
office (CTMO)
Collided
vehicles(intrusions)
Vehi cle
3
RSU RSU
Legend
Vehicles utilize inter-
vehicl es communication
and DSRC technology
Fog server
disseminate inter
vehicles
information and
DSRC technology
Figure 4. Proposed Secure Intelligent Vehicular Network using Fog Computing (SIVNFC) system
architecture prevention model.
Figure 4.
Proposed Secure Intelligent Vehicular Network using Fog Computing (SIVNFC) system
architecture prevention model.
The VSIF model deployment in SAPM includes the RSU connection with the FS Step 3. Scenario 3
(Figure 4) illustrates the deployment of VSIF model, the FS, and the RSU connection. FS collects intruder
Electronics 2019,8, 455 11 of 25
or collided vehicles or any unusual network attack information. The FS also obtains information
concerning all forms of DoS and SNI attacks that may be eminent in the network through the RSA
which is installed at the TMO. The TMO is presumed connected with the RSU. The VSIF model utilizes
inter-vehicle communications connections based upon the following deployment explanations.
RSU (Road side unit): RSUs are gateways. Gateways are also deployed in the proposed SAPM
which establishes connections with the FS. The RSU is equipped with network devices. It utilizes
DSRC inter-vehicle communication packet transfer based on IEEE 802.11.
RSU to FS: VANET utilizes V2V and V2RSU communication to propagate safety/non-safety
information. RSUs communicates with each other as well. Thus, RSU behaves as the FS backbone.
Wireless and wired connections are formed between RSU and FS (Figure 4). The RSU is aligned with FS.
Fog Server to Fog Server (FS to FS): FSs are identified at different locations. They interact with
each other. Consequently, a pool of VANET resources that is localized can be managed through the
TMO. This connection can be achieved via vehicular control center traffic management or TMO, as
shown in Figure 4. Thus, direct wireless and wired communication between peer FS can be possible.
In addition, collaborative services provision and the FS peer contents delivery can be initiated at the
TMO, which improves the entire SAPM. In addition, the cloud is logically connected with the FS and
has the tendency to aggregates information.
Fog Server to Cloud: In the proposed SAPM, FSs utilize fog computing to address location
awareness concern of cloud computing. Thus, cloud computing represents a central portal of
information which does not require location awareness for information processing. The cloud centrally
controls the FS in various locations. A FS possesses the capability to aggregate the information that
it has obtained from other FSs. The VSIF utilizes centralized computations whereby FS transmit
intervehicle information that it has received from the cloud to the application users [
36
], utilizing the
DSRC technology.
Due to open nature of the VANET deployment and associated vulnerabilities, RSU and the FS
utilize an authentication/KDE preventive mechanism in the proposed SAPM for ensuring real-time
packet delivery.
Electronics 2019, 8, x FOR PEER REVIEW 12 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
64123 65123
66123
67123
RSU
Fog server
Figure 5. VANET (vehicular ad hoc network) structure with Integrated Fog Server Model.
The proposed SAPM utilizes two levels of authentication/KDE preventive mechanisms for the
FS and the RSU aggregation of information, namely, the fog Level (FL) and the RSU Level (RSU-L).
The RSU-L considers the vehicle’s displacement and jitter in the VANET, whereas the FL utilizes the
Lagrange Polynomial for the identification of untrusted nodes as well [37].
3.5.1. FL Prevention Mechanism
The FL keeps one global key for the entire network; hence, each vehicle is identified by the global
key itself. Distributing the global key in the vehicles is insecure; therefore, the vehicles follow a shared
system. Each vehicle has its own shared value.
When a vehicle requests the information from a server either directly or through an RSU, the fog
server will demand three shares from any vehicle in the network or will choose two of them randomly
[38]. Three total shares will be considered, including the demanding vehicle. The fog server will
utilize the Lagrange polynomial to calculate the following.
The Lagrange polynomial S(X) containing degree ≤(n−1) demands n vehicles with
coordinates x,y=f(x),x,y=f(x), …… x,y=f(x) is given by:
S(X)=P
(X)
(1)
Where P is given by
P(X)=
y
x−x
x−x where1 ≥ 1, l ≤ n and l ! = k (2)
If written explicitly for n=3 vehicles,
S(X)=(x−x)(x−x)
(x−x)(x1−x3)
y
+ (x−x)(x−x)
(x−x)(x2− x3)
y
+ (x−x)(x−x)
(x−x)(x3− x2)
y
(3)
The separate polynomial can also be formulated as with Szeto (1975), which was later called
Lagrange’s fundamental interpolation.
Figure 5. VANET (vehicular ad hoc network) structure with Integrated Fog Server Model.
The proposed SAPM utilizes two levels of authentication/KDE preventive mechanisms for the
FS and the RSU aggregation of information, namely, the fog Level (FL) and the RSU Level (RSU-L).
Electronics 2019,8, 455 12 of 25
The RSU-L considers the vehicle’s displacement and jitter in the VANET, whereas the FL utilizes the
Lagrange Polynomial for the identification of untrusted nodes as well [37].
3.5.1. FL Prevention Mechanism
The FL keeps one global key for the entire network; hence, each vehicle is identified by the global
key itself. Distributing the global key in the vehicles is insecure; therefore, the vehicles follow a shared
system. Each vehicle has its own shared value.
When a vehicle requests the information from a server either directly or through an RSU, the
fog server will demand three shares from any vehicle in the network or will choose two of them
randomly [
38
]. Three total shares will be considered, including the demanding vehicle. The fog server
will utilize the Lagrange polynomial to calculate the following.
The Lagrange polynomial S(X) containing degree
≤(n−1)
demands
n
vehicles with coordinates
x1, y1=f(x1),x2, y2=f(x2),. . . . . . xn, yn=f(xn)is given by:
S(X)=
n
X
k=0
Pk(X)(1)
where Pkis given by
Pk(X)=yk
x−xl
xj−xl
where 1 ≥1, l ≤n and l !=k (2)
If written explicitly for n =3 vehicles,
S(X)=(x−x2)(x−x3)
(x1−x2)(x1 −x3)y1+(x−x1)(x−x3)
(x2−x1)(x2−x3)y2+(x−x1)(x−x2)
(x3−x1)(x3−x2)y3(3)
The separate polynomial can also be formulated as with Szeto (1975), which was later called
Lagrange’s fundamental interpolation.
S(X1)=x2∗x3
(x−x2)(x−x3)y1for the first vehicle (4)
S(X2)=x1∗x3
(x−x1)(x−x3)y2for the sec ond vehicle (5)
S(X3)=x1∗x2
(x−x2)(x−x3)y3for the third vehicle (6)
The key that is generated by the integration of separate polynomials is represented as
Gk=
n
X
k=0
S(k)(7)
If
Gk
matches the network key, only then does the vehicle pass any information from the fog server.
Second, the RSU level security is also applied, which makes the network more secure. To understand
the structure of this security, the pseudo code is also given as follows.
Electronics 2019,8, 455 13 of 25
Pseudo Code Algorithm for Share Verification
Notations:
SODFSV: Shares Ordering Demanded by FS from Vehicles:
ISVMyVALUE[]: Initial Share for Vehicle Value being Empty
SCV: Share for Current Vehicles
SKV: Share Key Value
SVCV: Share Vehicle Current value
Vi: Individual ith Number of Share for Vehicle
ICNSV: Initial Counter Number for Share of Vehicle
CSVBI: Current Share for Vehicle/Node Begin Iteration
SVNID :1st Share Key Vehicle/Node Identification or Initial Reference
SVNIDN um: Share Numerator key for Vehicle/Node Identification in Network
SVNIDDeno : Share Denominator key for Vehicle/Node Identification in Network
Vj: Individual jth Vehicle Chosen for Share in next Iteration
SVjC: When first Vehicle Share is Chosen there will be 2 remaining Share for the Vehicle
SVCNS: Share for Vehicle Chosen Current not same as next Share Chosen
RSCV: Remaining Share Counter for Current Vehicle
Input. S(k), n, i, k, j,
Process
1. Initialization
Vi=Vj;
ISVMyVALUE[ ] =∅;
SODFSV=2 ;
SVNIDDeno =∅;
2. If ISVMyV ALUE!=∅;
3. for Vi=1:3
While ICNSV = =1; then
a. CSVBI =SVNID;
b. for Vj== 1;
c. CSVBI =Vj;
d. If CSVBI !=Vj.
e. RSCV =SVjC;
4. RSCV=RSCV +1;
5. End if
6. End for
8. S VNIDDeno =Vj-RSCV ∗Vj−SV jC
9. SVNIDN um =RSCV ∗SVjC
10. ISVM MyVALUE[i]=SVNIDDeno
SVNIDNum
11. SKV=SVCV*ISVMyVALUE[i]
12. End for
Output :Gk, SCV
The pseudo code uses the interpolation order [
39
] of two and only three nodes for communication.
Whether the nodes will be selected for the data communication or not depends upon the final key
result, which is calculated using Lagrange’s method. One key generation method requires a numerator
and a denominator. The numerator is calculated using network IDs of the vehicles that remain for
the iteration [
40
]. For example, we consider 45, 53, and 61 to be the nodes that are selected for the
verification. Therefore, the numerator value
(Num)
for 45 is 53
∗
61
=
3233. The denominator
(deno)
is calculated by multiplying the difference of the network IDs of the remaining nodes. For 45, the
deno
value will be
(45 −53)∗(45 −61)→(−8)∗(−16)→128
. The verification key would be the product
of the Shared key of 45 to
Num
Deno
. Similarly, the
Shared key
for 53 and 61 will be calculated. The final
verification key would be the sum of all the generated verification keys.
Electronics 2019,8, 455 14 of 25
Finalkey =
i
X
k=0
Myvalue (8)
If the
Finalkey
is equal to the network security key, then the nodes are selected for communication.
Lagrange’s theorem randomly selects the nodes for verification. Though the verification process of
Lagrange is good enough, to make it more efficient, the CSA is applied to select the nodes for which
the verification keys will be generated. The CSA uses the node distance and its feedback to judge
whether it should be considered for key generation or not. Table 1shows the specifications considered
for the Cuckoo search algorithm (CSA).
LD =r((xnx1 −xnx2 )2+yny1 −yny22(9)
Table 1. The specifications considered for the Cuckoo search algorithm (CSA).
CSA Population Total Nodes in Coverage Region of Demanding Node
Fitness Parameters Feedback, Location Difference (LD)
LD is the location difference between the demanding node and the communicating node. The CSA
fetches the feedback values of nodes from the fog server, which also obtains intervehicle information
through the RSU.
As shown in Figure 6, the main node computes the distance between the demanding node and
the communicating node. It fetches the feedback from the fog server through the RSU and utilizes it
for the fitness function.
If Fitnessfunction →Return 1 ifd
fPn
k=0
dk
fk
n
Return 0 otherwise
(10)
where d is the distance between the fog and the user and f is the feedback of the fog server.
Electronics 2019, 8, x FOR PEER REVIEW 15 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
verification key would be the product of the Shared key of 45 to
. Similarly, the Shared for 53
and 61 will be calculated. The final verification key would be the sum of all the generated verification
keys.
Final =M
y
(8)
If the Final is equal to the network security key, then the nodes are selected for
communication. Lagrange’s theorem randomly selects the nodes for verification. Though the
verification process of Lagrange is good enough, to make it more efficient, the CSA is applied to select
the nodes for which the verification keys will be generated. The CSA uses the node distance and its
feedback to judge whether it should be considered for key generation or not. Table 1 shows the
specifications considered for the Cuckoo search algorithm (CSA).
Table 1. The specifications considered for the Cuckoo search algorithm (CSA).
CSA Population Total Nodes in Coverage Region of Demanding Node
Fitness Parameters Feedback, Location Difference (LD)
LD =
((x −x)+(y −y) (9)
LD is the location difference between the demanding node and the communicating node. The
CSA fetches the feedback values of nodes from the fog server, which also obtains intervehicle
information through the RSU.
As shown in Figure 6, the main node computes the distance between the demanding node and
the communicating node. It fetches the feedback from the fog server through the RSU and utilizes it
for the fitness function.
I
f
Fitness → Return 1 i
f
d
f
<
n (10)
Return 0 otherwise
where d is the distance between the fog and the user and f is the feedback of the fog server.
Communicating
vehicle
Fog server with
stored
feedback RSU
Commun icating
Distance d1
Demand ing
Vehi cle
C4 C3
C2
C1
d2
d3
d4
Feedbac k F
Figure 6. Node communication with a fog server.
The data transfer will take place once the route discovery process is complete. A network suffers
from two kinds of security issues—namely, the node level and the data level. This paper further
addresses the node level security [41].
Figure 6. Node communication with a fog server.
The data transfer will take place once the route discovery process is complete. A network suffers
from two kinds of security issues—namely, the node level and the data level. This paper further
addresses the node level security [41].
Electronics 2019,8, 455 15 of 25
3.6. Node level Security
VANET is a type of ad hoc network whose survival depends on vehicle/nodes cooperation and
trust. Therefore, trust between vehicles requires enforcement. Trust models can be categorized into
vehicle/node trust or data trust.
With node level trust security, vehicles/nodes evaluate trustworthiness between them, whereby
each vehicle crosscheck their neighbors redundant sensing data with their results. Trust in vehicles can
be calculated through a lightweight method and data which includes three parameters: Sensing a data
consistency value (or throughput), VANET communication ability, and the Vehicle/nodes remaining
lifetime. Trust assertion makes inconsistent data from DoS and SNI attacks to be detected [42].
The node level security is achieved by calculating the trust of neighboring nodes. The calculated
trust values are stored in the fog server for further processing.
The mathematical equation for node level security in the VANET is calculated by determining the
trust values of the node which is given as:
B=∓
n
X
i=1
Nxi(Y)(11)
The above equation shows that there is n number of trust factors. N(Y) indicates the trust value
of the node of ith category. It is seen that if B is greater than or equal to N, the associated risk is
less than threshold value and then node x will do work for Y. Node X keeps on checking to see any
recommendations about Y node from neighboring nodes, and, if so, the trust value is calculated using
the following equation.
C=Pz
x=1Nx(Y)
z(12)
where z indicates number of neighboring nodes and N
x
(Y) indicates the trust value of node X on
node Y.
The vehicles that have been identified as trusted nodes interact with the RSUs through the FS to
obtain the data in the appropriate order [
43
]. The proposed SIVNFC scheme utilizes an RSU prevention
mechanism whose model is as follows.
3.7. RSU-L Prevention Mechanism
The network deployment is based upon the specifications in Table 2.
Table 2. Network Specifications.
Total Number of Vehicles 50–100
Height of the Network 1000 m
Width of the Network 1000 m
Node Displacement 100–500 m/s
Simulation Iterations 1000
Simulation Tool MATLAB
Pseudo Code for Vehicle Placement
// To maintain the randomness in the network, the network is set in a random manner
1. For each n Nodes
2. Xloc(n)=1000*rand// Create a random x coordinate
3. Yloc(n)=1000*rand
4. lace(Xloc(n),Yloc(n))// Place the node in the network
5. End For
Electronics 2019,8, 455 16 of 25
Vehicles have different sets of parameters. The functions are designed to initiate the network
parameters. A real-time simulation may result in different structures. In addition, a network may not
include any fixed structure; however, for the sake of any simulation, some parameters should be initialized.
Pseudo Code to Initialize Vehicle Features
1. For i=1:Nodes // Loop running for each node
2. Delay_n(i)=Random D; // Include a delay value if the node is acting normally
3. Delay_t(i)=
Dealy_n2
;// For now, the expected reality is unpredictable; hence, just the random //architecture
is it set to be the square of the normal delay
4. End for
As the delay is initialized in a similar fashion, the other network parameters such as the jitter and
packet drop are also initialized. The battery consumption is not a problem in the case of a VANET
since the battery continues charging as long as the vehicle is running [44].
Figure 7a,b represents the path construction and attack mode of the attacker. Figure 7b shows that
the intensity of the attacker varies at different times. If the intensity is high, the attacker is attempting
to dump more packets. The above attacker scenario is demonstrated in the equations below.
Tpd =Pdn +Pda (13)
where Tpd is the total packet drop, Pdn is the total number of dropped packets in the normal mode,
and Pda is the dropped packets when the network is threatened.
Pdr =(Tp −Tpd)/Tp (14)
where Pdr is the packet delivery ratio, and Tp is the total number of packets. The random behavior of
an attack makes the network architecture more sophisticated. Now, the challenge is to identify them.
The proposed solution utilizes the feedforward back propagation neural network (FFBP-NN), and the
general utilities of the FFBP-NN are given in Table 3.
Electronics 2019, 8, x FOR PEER REVIEW 18 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
Vehicles have different sets of parameters. The functions are designed to initiate the network
parameters. A real-time simulation may result in different structures. In addition, a network may not
include any fixed structure; however, for the sake of any simulation, some parameters should be
initialized.
Pseudo Code to Initialize Vehicle Features
1. For i=1:Nodes // Loop running for each node
2. Delay_n(i)=Random D; // Include a delay value if the node is acting normally
3. Delay_t(i)= Dealy_n; // For now, the expected reality is unpredictable; hence, just the random
//architecture is it set to be the square of the normal delay
4. End for
As the delay is initialized in a similar fashion, the other network parameters such as the jitter
and packet drop are also initialized. The battery consumption is not a problem in the case of a VANET
since the battery continues charging as long as the vehicle is running [44].
Figures 7a,b represents the path construction and attack mode of the attacker. Figure 7b shows
that the intensity of the attacker varies at different times. If the intensity is high, the attacker is
attempting to dump more packets. The above attacker scenario is demonstrated in the equations
below.
Tpd = Pdn + Pda (13)
where Tpd is the total packet drop, Pdn is the total number of dropped packets in the normal mode,
and Pda is the dropped packets when the network is threatened.
Pdr = (Tp − Tpd)/Tp (14)
(a) Constructed path.
Figure 7. Cont.
Electronics 2019,8, 455 17 of 25
Electronics 2019, 8, x FOR PEER REVIEW 19 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
(b). Attacker.
Figure 7. (a) Constructed path; (b) attacker.
where Pdr is the packet delivery ratio, and Tp is the total number of packets. The random behavior
of an attack makes the network architecture more sophisticated. Now, the challenge is to identify
them. The proposed solution utilizes the feedforward back propagation neural network (FFBP-NN),
and the general utilities of the FFBP-NN are given in Table 3.
Table 3. Utilized feedforward back propagation neural network (FFBP-NN) structure.
Total Hidden Layer 1
Neuron Count 30
Feeding Iteration 100
Reverse Iteration 40–60
Propagation Type Linear
Algebraic Model Levenberg
The Artificial Intelligence (AI) method is made up of two sections:
• Training and Classification
The classification section is used in the identification model. The training module utilizes the
jitter as the training parameter. To train the neural network, the neural network toolbox in MATLAB
is utilized. The training layer is provided with the target set as well. The target is the identification of
the nodes. The training consists of two phases. First, training is performed for the identification of
the path, and then the training is performed for the identification of the affected vehicle(s) in the route
[45].
The following equation can be defined:
Jtr=Dp (a, n) +Nd (15)
where Jtr is the jitter, Dp is the delay of the path, and ‘a’ and ‘n’ represent the advanced (under threat)
and normal situations, respectively. Nd is the network delay. For each path in every iteration, there
will be jitter. The proposed solution uses the first 400 iterations’ data for training and then uses the
next 600 iterations’ data with the training structure for identification.
Figure 7. (a) Constructed path; (b) attacker.
Table 3. Utilized feedforward back propagation neural network (FFBP-NN) structure.
Total Hidden Layer 1
Neuron Count 30
Feeding Iteration 100
Reverse Iteration 40–60
Propagation Type Linear
Algebraic Model Levenberg
The Artificial Intelligence (AI) method is made up of two sections:
Training and Classification
The classification section is used in the identification model. The training module utilizes the
jitter as the training parameter. To train the neural network, the neural network toolbox in MATLAB is
utilized. The training layer is provided with the target set as well. The target is the identification of the
nodes. The training consists of two phases. First, training is performed for the identification of the
path, and then the training is performed for the identification of the affected vehicle(s) in the route [
45
].
The following equation can be defined:
Jtr =Dp (a, n) +Nd (15)
where Jtr is the jitter, Dp is the delay of the path, and ‘a’ and ‘n’ represent the advanced (under threat)
and normal situations, respectively. Nd is the network delay. For each path in every iteration, there
will be jitter. The proposed solution uses the first 400 iterations’ data for training and then uses the
next 600 iterations’ data with the training structure for identification.
Electronics 2019,8, 455 18 of 25
Algo Train_Neural (Iteration_Data,Total_Iterations)
For i=1:Total_Iterations
Training Data (i) =Iteration Data (i);
Targetable (i) =Path ID;
End For
Neural=Initialize Neural (Training Data, Target Label, k); // k→Total Neurons (30 in this case)
NeuralI.TrainParam.Epochs=100; // Total training iterations
Train (NeuralITraining_Data, Target_Label); // Training with Initialized Neural and Training data
End Algorithm
The training section results in FFBP-NN structure given in Figure 8.
Electronics 2019, 8, x FOR PEER REVIEW 20 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
Algo Train_Neural (Iteration_Data,Total_Iterations)
For i=1:Total_Iterations
Training Data (i) =Iteration Data (i);
Targetable (i) =Path ID;
End For
Neural=Initialize Neural (Training Data, Target Label, k); // k Total Neurons (30 in this case)
NeuralI.TrainParam.Epochs=100; // Total training iterations
Train (NeuralITraining_Data, Target_Label); // Training with Initialized Neural and
Training data
End Algorithm
The training section results in FFBP-NN structure given in Figure 8.
(a) Feed forward structure.
(b) Back propagation firefly.
Figure 8. (a) Feed Forward Structure; (b) Back Propagation Firefly.
Figure 8. (a) Feed Forward Structure; (b) Back Propagation Firefly.
Electronics 2019,8, 455 19 of 25
3.8. Identification of Affected Node(s) and Recovery
The proposed research work also presents a regression model with backpropagation. Figure 9
represents the regression model and values.
Electronics 2019, 8, x FOR PEER REVIEW 21 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
3.8. Identification of Affected Node(s) and Recovery
The proposed research work also presents a regression model with backpropagation. Figure 9
represents the regression model and values.
Figure 9. Regression model.
4. Results and Analysis
4.1. Feed Forward–Backward Propagation and Regression Model Result and Analysis
4.1.1. Feed Forward–Backward Propagation
From Figure 8, we can see that the proposed SIVNFC scheme calculates both the training data
for latency (jitter) and validation of jitter that is the deviation between the predicted y and the actual
y as a measure by the mean squared error (MSE). We can see that we have five Epochs for our model.
This means that we are essentially training our model over five forwards and backwards. The five
epoch is also the stopping iteration and the one epoch for back iteration. The expectation is that the
proposed SIVNFC scheme will decrease with each epoch, which means that our model is predicting
the value of y more accurately as we continue to train the model.
The predictions of the test data show how good the proposed SIVNFC scheme is. The test graph
in Figure 8b, which indicates validation performance at epoch 1 of the model, indicates our model
predictions is a good one.
From the graph in Figure 8b, we can see that both the training and the validation loss decreases
in exponential fashion as the number of epochs is increased. This suggests that the model has gained
high degree of accuracy as our epochs (i.e., the number of forward and backward passes) is increased.
4.1.2. Regression Model Result and Analysis
Figure 9 represents the close and high regression value of the proposed scheme. The result
indicates that the proposed model close and high regression values are: Training is 0.98748, validation
is 0.97053, test is 0.97357, and the value for all is 0.98209. All these regression values are close and
high as well. Close and high regression values generally represent healthy training and classification
structure. High regression value is the reason because of which the prevention parameters are high
for the proposed model to prevent much jitter/delays in the SAPM architecture.
As discussed earlier, this section classifies the path value on the basis of the trained structure.
The identified attacker nodes are always sent for recovery or maintenance.
Figure 9. Regression model.
4. Results and Analysis
4.1. Feed Forward–Backward Propagation and Regression Model Result and Analysis
4.1.1. Feed Forward–Backward Propagation
From Figure 8, we can see that the proposed SIVNFC scheme calculates both the training data for
latency (jitter) and validation of jitter that is the deviation between the predicted y and the actual y
as a measure by the mean squared error (MSE). We can see that we have five Epochs for our model.
This means that we are essentially training our model over five forwards and backwards. The five
epoch is also the stopping iteration and the one epoch for back iteration. The expectation is that the
proposed SIVNFC scheme will decrease with each epoch, which means that our model is predicting
the value of y more accurately as we continue to train the model.
The predictions of the test data show how good the proposed SIVNFC scheme is. The test graph
in Figure 8b, which indicates validation performance at epoch 1 of the model, indicates our model
predictions is a good one.
From the graph in Figure 8b, we can see that both the training and the validation loss decreases in
exponential fashion as the number of epochs is increased. This suggests that the model has gained
high degree of accuracy as our epochs (i.e., the number of forward and backward passes) is increased.
4.1.2. Regression Model Result and Analysis
Figure 9represents the close and high regression value of the proposed scheme. The result
indicates that the proposed model close and high regression values are: Training is 0.98748, validation
is 0.97053, test is 0.97357, and the value for all is 0.98209. All these regression values are close and
high as well. Close and high regression values generally represent healthy training and classification
structure. High regression value is the reason because of which the prevention parameters are high for
the proposed model to prevent much jitter/delays in the SAPM architecture.
As discussed earlier, this section classifies the path value on the basis of the trained structure.
The identified attacker nodes are always sent for recovery or maintenance.
Electronics 2019,8, 455 20 of 25
4.2. QoS Provision Analysis in VANET
Development of VANET has recently received attention. Most of these attentions were based on
the research effort conducted in the industry and in the field of academia [
46
]. VANET is classified as a
key technology in intelligent transportation systems. VANET is envisaged as playing an important
role in the futuristic smart cities. This important role in VANET improves road safety and also
provide innovative services relating to traffic management and information achievement applications.
Thus, it has become expedient for creating a wide range of services for future VANET deployment
that ranges from safety/security and traffic management to commercial applications services [
47
].
Offering these services requires high QoS guarantees. Without QoS guarantees, these services would
not be successfully achieved. Due to the highly dynamic nature of VANET, resources reservation for
services are not applicable for providing a QoS guarantee.
In addition, two communicating vehicles that are moving would experience a degrading
performance. This can be possible when the wireless links formed between them are vulnerable and
the vehicles are disconnected due to DoS attacks. This can lead to unpredictable driver performance.
QoS metrics such as throughput and jitter associated with the current routes established changes
rapidly. The best selected routes computed by the RSU could easily become inefficient and lead to
infeasible routes due to imminent links breakdown. Thus, utilizing a search for feasible route in
multihop VANET is subject to multiple QoS constraints.
4.2.1. QoS Results and Analysis for the Proposed Model
The result and analysis of the proposed SIVNFC scheme is compared with the other contending
models such as: CSA (Cuckoo), FA (firefly), and the firefly neural network. The analysis is based upon
the QoS provision determination in VANET. The QoS analysis is based upon the simulation result and
the mathematical analysis of the models in the SAPM. The QoS investigation is centered on throughput
and jitter associated with the currents routes that has been established in the network as a result of
rapid changes in the network due to the result of DoS and SNI in the VANET. We determine the QoS
as follows:
•Throughput: It is the total number of delivered packets in the given time frame.
Throughput =Totaldelivered
Timeframe
(16)
•Latency/jitter: It is the total delay that is produced when delivering data packets in the network.
The evaluation of the parameters is obtained in such a manner that the Packet Injection Rate (PIR)
is on the x-axis and the QoS evaluation parameter is on the y-axis. The PIR is the ratio of the injection
of the packets into the network.
Figure 10 demonstrates the results of the proposed SVINFC scheme, which is compared with all
the other contending models. The proposed SINVNFC scheme considers the throughput with Cuckoo,
firefly, and the firefly neural network. The range of
PIR
is from 0.001 to 0.02. With the increase in the
PIR, the throughput increases, which is also demonstrated in Figure 10. The maximum throughput at
PIR
=0.02 is 8100 for the proposed
SIVNFC
scheme and 7900 for the firefly–neural network odel. One
hundred packets are injected per millisecond.
Electronics 2019,8, 455 21 of 25
Electronics 2019, 8, x FOR PEER REVIEW 23 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
Figure 10. Throughput versus PIR.
The second evaluation parameter is the jitter . Jitter produces delays when the network
experiences DoS and SNI. However, due to the fact that the proposed SVINFC scheme has introduced
fog computing and that trust between the communicating neighboring nodes has been established,
the entire network level security is increased. This has also led to decreased communication costs and
time. The route that is discovered and assigned as trusted is stored on the fog server. Due to this, the
need for broadcasting is reduced for route discovery and much time is saved. The evaluation of the
jitter is done considering the same aspects as the throughput.
The jitter is not a consistent parameter in any network. Figure 11 shows that the jitter may be
high or low for different PIR values. Throughout the PIR, the proposed SIVNFC scheme is noted to
produce the least jitter when compared to other contending models scenarios. Though the fog
computing server is applied to all the scenarios, the max jitter for the SIVNFC scheme is
96 ms, whereas the maximum jitter for Firefly Neural is 102 ms.
Figure 11. Jitter versus Packet Injection Rate (PIR).
Figure 12 represents close but least/high regression values of the proposed scheme. These results
show detail regression model that was generated in the simulation before the final regression values
were obtained. The result generated includes the following: The training result is 0.97847, the
validation result is NaN (not a number), the test result is NaN, and the value for all result is 0.98727.
Close and least/high regression values generally represent healthy training and classification
structure as well, as indicated previously.
0
100
200
0.0010.0020.0060.008 0.01 0.0120.0140.018 0.02
JITTER IN ms
PIR
jitter
SIVNFC Jitter Firefly-Neural
Jitter Cuckoo Jitter Firefly
Figure 10. Throughput versus PIR.
The second evaluation parameter is the
jitter
. Jitter produces delays when the network experiences
DoS and SNI. However, due to the fact that the proposed SVINFC scheme has introduced fog computing
and that trust between the communicating neighboring nodes has been established, the entire network
level security is increased. This has also led to decreased communication costs and time. The route that
is discovered and assigned as trusted is stored on the fog server. Due to this, the need for broadcasting
is reduced for route discovery and much time is saved. The evaluation of the jitter is done considering
the same aspects as the throughput.
The
jitter
is not a consistent parameter in any network. Figure 11 shows that the
jitter
may be high
or low for different PIR values. Throughout the
PIR
, the proposed
SIVNFC
scheme is noted to produce
the least
jitter
when compared to other contending models scenarios. Though the fog computing server
is applied to all the scenarios, the max
jitter
for the
SIVNFC
scheme is 96
ms
, whereas the maximum
jitter for Firefly Neural is 102 ms.
Electronics 2019, 8, x FOR PEER REVIEW 23 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
Figure 10. Throughput versus PIR.
The second evaluation parameter is the jitter . Jitter produces delays when the network
experiences DoS and SNI. However, due to the fact that the proposed SVINFC scheme has introduced
fog computing and that trust between the communicating neighboring nodes has been established,
the entire network level security is increased. This has also led to decreased communication costs and
time. The route that is discovered and assigned as trusted is stored on the fog server. Due to this, the
need for broadcasting is reduced for route discovery and much time is saved. The evaluation of the
jitter is done considering the same aspects as the throughput.
The jitter is not a consistent parameter in any network. Figure 11 shows that the jitter may be
high or low for different PIR values. Throughout the PIR, the proposed SIVNFC scheme is noted to
produce the least jitter when compared to other contending models scenarios. Though the fog
computing server is applied to all the scenarios, the max jitter for the SIVNFC scheme is
96 ms, whereas the maximum jitter for Firefly Neural is 102 ms.
Figure 11. Jitter versus Packet Injection Rate (PIR).
Figure 12 represents close but least/high regression values of the proposed scheme. These results
show detail regression model that was generated in the simulation before the final regression values
were obtained. The result generated includes the following: The training result is 0.97847, the
validation result is NaN (not a number), the test result is NaN, and the value for all result is 0.98727.
Close and least/high regression values generally represent healthy training and classification
structure as well, as indicated previously.
0
100
200
0.0010.0020.0060.008 0.01 0.0120.0140.018 0.02
JITTER IN ms
PIR
jitter
SIVNFC Jitter Firefly-Neural
Jitter Cuckoo Jitter Firefly
Figure 11. Jitter versus Packet Injection Rate (PIR).
Figure 12 represents close but least/high regression values of the proposed scheme. These results
show detail regression model that was generated in the simulation before the final regression values
were obtained. The result generated includes the following: The training result is 0.97847, the validation
result is NaN (not a number), the test result is NaN, and the value for all result is 0.98727. Close and
least/high regression values generally represent healthy training and classification structure as well, as
indicated previously.
Electronics 2019,8, 455 22 of 25
Electronics 2019, 8, x FOR PEER REVIEW 24 of 27
Electronics 2019, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/electronics
Figure 12. Detail regression model generated during simulations.
5. Conclusion
This paper proposed a fog-integrated VANET scheme termed SIVNFC. The proposed scheme
simultaneously considers the node level and network level security. The node level security includes
the fog computing merged with the VANET, a node level security mechanism, a new fitness function
of the Cuckoo search, and a collaborated neural network structure. The node level security establishes
trust collaboration with all the neighbors of the network. The node level trustworthiness ensures that
the entire network rapidly delivers packets to the entire network system. The proposed SIVNFC
scheme also prevent DoS attacks and SNI from attacking the entire network. The proposed SIVNFC
scheme is an ad hoc network, and new vehicles, including DoS attacks, may easily enter into the
network. To prevent the network from being accessed by foreigners (or outsiders) until they become
part of the home network, the proposed SIVNFC scheme uses Lagrange’s interpolation method
through which node level and the network level security attacks such as DoS attacks entry is secured.
The proposed SIVNFC scheme also utilizes an integrated SAPM. The SAPM includes
intrusion/attacker and VSIF models. Both models deployment in SAPM are utilized to mitigate all
other forms of attacks and secure the network. The models are also deployed to provide real-time
information in the network through safety application deployment of the RSU at the TMO, where
information can be processed on timely to reduce delay and enhanced the throughput in the network.
The evaluation of the proposed SIVNFC scheme is evaluated using QoS parameters—namely, the
throughput and jitter. The proposed model is also compared with the firefly algorithm, a single
neural network, a neural network combined with the firefly algorithm, and the Cuckoo Search
algorithm.
The evaluation of the QoS parameters is done using the PIR as the basis of every simulation. The
proposed SIVNFC scheme provided a total throughput of 8100 for the PIR value of 0.2. The maximum
throughput of the network was also offered. For the same scenario, the second-best throughput was
7900 for the combination of Firefly and the neural network. The jitter is inconsistent throughout the
simulations, and it varied based on the model architecture and algorithm. Even after nonlinear
computations, the jitter for the SIVNFC scheme is a maximum of 96 ms, whereas it is 102 ms for the
firefly neural network.
The proposed SIVNFC scheme also utilizes the regression model to indicate the reduced delay
of the network. The current research work has potential for future research directions. The neural
Figure 12. Detail regression model generated during simulations.
5. Conclusions
This paper proposed a fog-integrated VANET scheme termed SIVNFC. The proposed scheme
simultaneously considers the node level and network level security. The node level security includes
the fog computing merged with the VANET, a node level security mechanism, a new fitness function
of the Cuckoo search, and a collaborated neural network structure. The node level security establishes
trust collaboration with all the neighbors of the network. The node level trustworthiness ensures
that the entire network rapidly delivers packets to the entire network system. The proposed SIVNFC
scheme also prevent DoS attacks and SNI from attacking the entire network. The proposed SIVNFC
scheme is an ad hoc network, and new vehicles, including DoS attacks, may easily enter into the
network. To prevent the network from being accessed by foreigners (or outsiders) until they become
part of the home network, the proposed SIVNFC scheme uses Lagrange’s interpolation method through
which node level and the network level security attacks such as DoS attacks entry is secured.
The proposed SIVNFC scheme also utilizes an integrated SAPM. The SAPM includes
intrusion/attacker and VSIF models. Both models deployment in SAPM are utilized to mitigate
all other forms of attacks and secure the network. The models are also deployed to provide real-time
information in the network through safety application deployment of the RSU at the TMO, where
information can be processed on timely to reduce delay and enhanced the throughput in the network.
The evaluation of the proposed SIVNFC scheme is evaluated using QoS parameters—namely, the
throughput and jitter. The proposed model is also compared with the firefly algorithm, a single neural
network, a neural network combined with the firefly algorithm, and the Cuckoo Search algorithm.
The evaluation of the QoS parameters is done using the PIR as the basis of every simulation.
The proposed SIVNFC scheme provided a total throughput of 8100 for the PIR value of 0.2.
The maximum throughput of the network was also offered. For the same scenario, the second-best
throughput was 7900 for the combination of Firefly and the neural network. The jitter is inconsistent
throughout the simulations, and it varied based on the model architecture and algorithm. Even after
nonlinear computations, the jitter for the SIVNFC scheme is a maximum of 96 ms, whereas it is 102 ms
for the firefly neural network.
The proposed SIVNFC scheme also utilizes the regression model to indicate the reduced delay
of the network. The current research work has potential for future research directions. The neural
Electronics 2019,8, 455 23 of 25
network structure can be varied to assess if there are any differences in the QoS parameters. A hybrid
classifier can also be tested to see if it enhances the current proposed neural architecture. This paper
utilized Lagrange’s interpolation method, and it would be interesting to examine the performances of
other interpolation methods such as Spline and the Polynomial fit. A combination of interpolation
methods can also be considered.
Author Contributions:
Conceptualization, S.K.E. and K.M.E.; Writing—original draft, S.K.E.; Writing—review
editing, K.E.
Funding: This research has been funded by University of Bridgeport.
Acknowledgments:
This research was developed and conducted by Samuel Kofi Erskine and Khaled M. Elleithy.
The algorithms and formulas were solely developed by the authors. This research is sponsored by the University
of Bridgeport.
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
Singh, D.; Ranvijay; Yadav, R.S. A state-of-art approach to misbehaviour detection and revocation in VANET:
Survey. Int. J. Ad Hoc Ubiquitous Comput. 2018,28, 77–93. [CrossRef]
2.
Cooper, C.; Franklin, D.; Ros, M.; Safaei, F.; Abolhasan, M. A comparative survey of VANET clustering
techniques. IEEE Commun. Surv. Tutor. 2017,19, 657–681. [CrossRef]
3.
Hasrouny, H.; Samhat, A.E.; Bassil, C.; Laouiti, A. VANET security challenges and solutions: A survey.
Veh. Commun. 2017,7, 7–20. [CrossRef]
4.
Sharma, S.; Kaul, A. A survey on Intrusion Detection Systems and Honeypot based proactive security
mechanisms in VANETs and VANET Cloud. Veh. Commun. 2018,12, 138–164. [CrossRef]
5.
Panayappan, R.; Trivedi, J.M. VANET-Based Approach for Parking Space Availability; Carnegie Cylab Mellon
University: Pittsburgh, PA, USA, 2007; pp. 1–4.
6.
Grover, J.; Jain, A.; Singhal, S.; Yadav, A. Real-Time VANET Applications Using Fog Computing.
In Proceedings of the First International Conference on Smart System, Innovations and Computing, Smart
Innovation, Systems and Technologies, Jaipur, India, 15–16 April 2017; Springer Nature: Singapore, 2018;
pp. 685–687.
7.
Glass, S.; Mahgoub, I.; Rathod, M. Leveraging MANET-Based Cooperative Cache Discovery Techniques in
VANETs: A Survey and Analysis. IEEE Commun. Surv. Tutor. 2017,19, 2640–2661. [CrossRef]
8.
Chaba, S.; Kumar, R.; Pant, R.; Dave, M. Secure and efficient key delivery in VANET using cloud and
fog computing. In Proceedings of the 2017 International Conference on Computer, Communications and
Electronics (Comptelix), Jaipur, India, 1–2 July 2017; pp. 27–31.
9.
Calandrelli, G.; Papadimitratos, P.; Hubaux, J.-P.; Lioy, A. Efficient and Robust Pseudonymous Authentication in
VANET; Laboratory for Computer Communications and Applications, EPFL: Lausanne Switzerland, 2007;
pp. 19–23.
10.
Li, R.; Jin, L. Improved Cuckoo Algorithm for Spectrum Allocation in Cognitive Vehicular Network.
In Proceedings of the 2018 5th International Conference on Systems and Informatics (ICSAI 2018), Nanjing,
China, 10–12 November 2018; pp. 823–833.
11.
Zhang, R.; Jiang, X.; Li, R. Decomposition based multiobjective spectrum allocation algorithm for cognitive
vehicular networks. In Proceedings of the 2017 IEEE 17th International Conference on Communication
Technology (ICCT), Chengdu, China, 27–30 October 2017; pp. 1–3.
12.
Narawade, V.E.; Kolekar, U.D. EACSRO: Epsilon constraint-based Adaptive Cuckoo Search algorithm for
Rate Optimized Congestion Avoidance and Control in Wireless Sensor Networks. In Proceedings of the
2017 International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC), Palladam,
India, 10–11 February 2017; pp. 715–720.
13.
Azmat, F.; Chen, Y.; Stocks, N. Analysis of Spectrum Occupancy Using Machine Learning Algorithms.
IEEE Trans. Veh. Technol. 2016,65, 6853–6854. [CrossRef]
14.
Sachdev, A.; Mehta, K.; Malik, L. Design of Protocol for cluster based routing in VANET using Fire Fly
Algorithm. In Proceedings of the 2016 IEEE International Conference on Engineering and Technology
(ICETECH), Coimbatore, India, 17–18 March 2016; pp. 1–3.
Electronics 2019,8, 455 24 of 25
15.
Kumar, R.; Chhabra, S. Efficient routing in Vehicular Ad-hoc Networks using Firefly optimization.
In Proceedings of the 2016 International Conference on Inventive Computation Technologies (ICICT),
Coimbatore, India, 26–27 August 2016; pp. 1–6.
16.
Ghaleb, F.A.; Zainal, A.; Rassam, M.A.; Mohammed, F. An effective misbehavior detection model using
artificial neural network for vehicular ad hoc network applications. In Proceedings of the2017 IEEE
Conference on Application, Information and Network Security (AINS), Miri, Malaysia, 13–14 November
2017; pp. 1–5.
17.
Agarwal, Y.; Jain, K.; Karabasoglu, O. Smart vehicle monitoring and assistance using cloud computing in
Vehicular Ad Hoc networks. Int. J. Transp. Sci. Technol. 2018,7, 60–73. [CrossRef]
18.
Farhan, A.; Muhammad, K.; Asma, A.; Abir, A. Vehicular Cloud Networks Architecture, Applications and
Security Issues. In Proceedings of the 2015 IEEE/ACM. International Conference on Utility and Cloud
Computing (UCC), Limassol, Cyprus, 7–10 December 2015; pp. 571–576.
19.
Vennila, R.; Duraisamy, V. Inter cluster communication and rekeying technique for multicast security in
mobile ad hoc networks. IET Inf. Secur. 2014,8, 234–239.
20.
Kang, J.; Lin, D.; Jiang, W.; Bertino, E. Highly efficient randomized authentication in VANETs. Pervasive Mob.
Comput. 2018,44, 31–44. [CrossRef]
21.
Dalya, K.; Omprakash, S.; Kaiwartya, O.; Abdullah, A.H.; Hassan, N. Location Information Verification cum
Security using TBM in Geocast Routing. In Proceedings of the International Conference on Eco-Friendly
Computing and Communication Systems, Haryana, India, 7–8 December 2015; pp. 219–221.
22.
Qureshi, K.N.; Abdullah, A.H.; Kaiwartya, O.; Iqbal, S.; Butt, R.A.; Bashir, F. A Dynamic Congestion Control
Scheme for safety applications in vehicular ad hoc networks. Comput. Electr. Eng.
2018
,72, 774–788.
[CrossRef]
23.
Kasana, R.; Kumar, S.; Kaiwartya, O.; Yan, W.; Cao, Y.; Abdullah, A.H. Location error resilient geographical
routing for vehicular ad-hoc networks. IET Intell. Transp. Syst. 2017,11, 450–452. [CrossRef]
24.
Dora, D.P.; Kumar, S.; Kaiwartya, O.; Prakash, S. Secured Time Stable Geocast (S-TSG) routing for VANET.
In Proceedings of the International Conference on Advanced Computing, Networking and Informatics,
Smart Innovation, Systems and Technologies 2015, Bhubaneswar, India, 23–25 June 2015; pp. 161–162.
25.
Kaiwartya, O.; Cao, Y.; Lloret, J.; Kumar, S.; Aslam, N.; Kharel, R.; Abdullah, A.H.; Shah, R.R. Geometry-based
Localization for GPS Outage in Vehicular Cyber Physical Systems. IEEE Trans. Veh. Technol.
2018
,67,
3800–3801. [CrossRef]
26.
Tangade, S.; Manvi, S.S.; Lorenz, P. Decentralized and Scalable Privacy-Preserving Authentication Scheme in
VANETs. IEEE Trans. Veh. Technol. 2018,67, 8647–8655. [CrossRef]
27.
Lai, Y.; Zhang, L.; Wang, T.; Yang, F.; Xu, Y. Data Gathering Framework Based on Fog Computing Paradigm in
VANETs. In Proceedings of the Asia-Pacific Web (APWeb) and Web-Age Information Management (WAIM)
Joint Conference on Web and Big Data, Macau, China, 23–25 July 2017; Springer: Cham, Switzerland, 2017;
pp. 227–236.
28.
Shehab, M.; Khader, A.T.; Al-Betar, M.A. A survey on applications and variants of the cuckoo 797 search
algorithm. Appl. Soft Comput. 2017,61, 1041–1059. [CrossRef]
29.
Liao, D.; Li, H.; Sun, G.; Zhang, M.; Chang, V. Location and trajectory privacy preservation in 5G-Enabled
vehicle social network services. J. Netw. Comput. Appl. 2018,110, 108–118. [CrossRef]
30.
Singh, P.; Bart, W.N. Prevention of denial of service attack over vehicle ad hoc network using quick response
table. In Proceedings of the 2017 4th International Conference on Signal Processing, Computing and Control
(ISPCC), Solan, India, 21–23 September 2017; pp. 568–589.
31.
Wang, M.; Liang, H.; Deng, R.; Zhang, R.; Shen, X.S. VANET based online charging strategy for electric
vehicles. In Proceedings of the 2013 IEEE Global Communications Conference (GLOBECOM), Atlanta, GA,
USA, 9–13 December 2013; pp. 4804–4809.
32.
Giulio, M.; Lorenzo, V. On the performance of channel occupancy detectors for vehicular ad-hoc networks.
In Proceedings of the 2013 IEEE 5th International Congress on Ultra Modern Telecommunications and
Control Systems and Workshops (ICUMT), Almaty, Kazakhstan, 10–13 September 2013; pp. 1–6.
33.
Pathre, A.; Agrawal, C.; Jain, A. A novel defense scheme against DDOS attack in VANET. In Proceedings of
the 2013 IEEE Tenth International Conference on Wireless and Optical Communications Networks (WOCN),
Bhopal, India, 26–28 July 2013; pp. 1–5.
Electronics 2019,8, 455 25 of 25
34.
Bitam, S.; Mellouk, A. Bee life-based multi constraints multicast routing optimization for vehicular ad hoc
networks. J. Netw. Comput. Appl. 2013,36, 981–991. [CrossRef]
35.
Pathre, A. Identification of malicious vehicle in vanet environment from DDOS attack. J. Glob. Res. Comput. Sci.
2013,4, 30–34.
36.
Whaiduzzaman, M.; Sookhak, M.; Gani, M. A survey on Vehicular cloud computing. J. Netw. Comput. Appl.
2014,40, 325–344. [CrossRef]
37.
Liu, J.; Li, J.; Zhang, L.; Dai, F.; Zhang, Y.; Meng, X.; Shen, J. Secure intelligent traffic light control using fog
computing. Future Gener. Comput. Syst. 2018,78, 817–824. [CrossRef]
38.
Sookhak, M.; Yu, F.R.; Tang, H. Secure data sharing for vehicular ad-hoc networks using cloud computing.
In Proceedings of the Ad Hoc Networks, Ottawa, ON, Canada, 28–30 June 2017; Springer: Cham, Switzerland,
2017; pp. 306–315.
39.
Nobre, J.C.; de Souza, A.M.; Ros
á
rio, D.; Both, C.; Villas, L.A.; Cerqueira, E.; Gerla, M. Vehicular
software-defined networking and fog computing: Integration and design principles. Ad Hoc Netw.
2019
,82,
172–181. [CrossRef]
40.
Rauniyar, A.; Hagos, D.H.; Shrestha, M. A Crowd-Based Intelligence Approach for Measurable Security,
Privacy, and Dependability. Internet of Automated Vehicles with Vehicular Fog. Mob. Inf. Syst.
2018
,2018,
828–829. [CrossRef]
41.
He, Y.; Wei, Z.; Du, G.; Li, J.; Zhao, N.; Yin, H. Securing Cognitive Radio Vehicular Ad hoc Networks with
Fog Computing. Ad Hoc Sens. Wirel. Netw. 2018, 1–4.
42.
Anbuchelian, S.; Lokesh, S.; Baskaran, M. Improving Security In Wireless Sensor Network Using TrustAnd
Metaheuristic Algorithms. In Proceedings of the 2016 3rd International Conference on Computer and
Information 859 Sciences (ICCOINS), Kuala Lumpur, Malaysia, 15–17 August 2016; pp. 233–238.
43.
Wei, J.; Wang, X.; Li, N.; Yang, G.; Mu, Y. A Privacy-Preserving Fog Computing Framework for Vehicular
Crowd sensing Networks. IEEE Access 2018,6, 43776–43784. [CrossRef]
44.
Islam, S.H.; Obaidat, M.S.; Vijayakumar, P.; Abdulhay, E.; Li, F.; Reddy, M.K.C. A robust and efficient
password-based conditional privacy preserving authentication and group-key agreement protocol for
VANETs. Future Gener. Comput. Syst. 2018,84, 216–227. [CrossRef]
45.
Sundarasekar, R.; Thanjaivadivel, M.; Manogaran, G.; Kumar, P.M.; Varatharajan, R.; Chilamkurti, N.;
Hsu, C.H. Internet of Things with Maximal Overlap Discrete Wavelet Transform for Remote 839 Health
Monitoring of Abnormal ECG Signals. J. Med. Syst. 2018,42, 228. [CrossRef] [PubMed]
46.
Yu, R.; Zhang, Y.; Gjessing, S.; Xia, W.; Yang, K. Toward cloud-based vehicular networks with efficient
resource management. IEEE Netw. 2013,27, 48–54. [CrossRef]
47.
Eiza, M.H.; Owens, T.; Ni, Q.; Shi, Q. Situation-Aware QoS Routing Algorithm for Vehicular Ad Hoc
Networks. IEEE Trans. Veh. Technol. 2015,64, 5520–5522. [CrossRef]
©
2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
