Wireless Networks (2023) 29:2551–2575

https://doi.org/10.1007/s11276-023-03336-2 (0123456789().,-volV)(0123456789().
,- volV)

ORIGINAL PAPER

ReTrust: reliability and recommendation trust-based scheme for secure

data sharing among internet of vehicles (IoV)
Kuldeep Narayan Tripathi1,2 • Ashish Mohan Yadav3 • Surendra Nagar4 • S. C. Sharma2

Accepted: 27 March 2023 / Published online: 16 April 2023

 The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023

Abstract
The Internet of Vehicles (IoV) uses wireless communication for establishing a distributed, dynamic, and temporary
network among the vehicles on the road. Due to their dynamic topology, distributed structure, and open nature, vehicular
communication networks (VCN) are vulnerable to numerous security vulnerabilities. Moreover, the limited transmission
range of vehicles and the dynamic network topology makes data transmission arduous and require a special routing
mechanism to handle these challenges in real-time. This work aims to utilize reliable intermediate relay vehicles to
securely forward messages to the appropriate recipients in vehicular communication networks. A trust-based security
approach has been fascinating in a vehicular communication network to identify rogue vehicles and establish a reliable
communication channel among the vehicles. This paper presents a new reliability and recommendations (ReTrust) based
scheme using a combined trust model to detect rogue nodes in the IoV network. A convenient combined trust-based
security algorithm has been proposed to assess vehicles’ trustworthiness and identify the rogue vehicles in the network
based on the estimated trust value. Moreover, Direct trust has been calculated based on multiple factors, including
communication and data reliability. In addition, the recommendation trust has been assessed using the cooperation of
neighbouring vehicles. The proposed method uses the similarity measure and correlation coefficient to classify the fab-
ricated recommendation values in the network. Extensive simulation tests were conducted using the network simulator NS-
2 and VANET-MobiSim to evaluate the suggested method’s performance. The effectiveness of the proposed scheme has
been evaluated using different quality of service parameters, and the results reveal that the ReTrust performs admirably.
The simulation outcomes validate that the ReTrust scheme has a high detection rate ([ 90%) and a low false-positive rate
(\ 5%). This mechanism could be useful for creating a trusted vehicular network and providing secure vehicle commu-
nication for safety-related applications.

Keywords Trust  Rogue Nodes  Recommendation Trust  Relay Vehicles  VANETs  Secure Routing

1 Introduction

The rapid development in the current road transportation

& Kuldeep Narayan Tripathi scenario makes a revolution in the vehicular intelligent
[email protected]
communication system. In recent years, several works have
1
Department of Computer Science and Engineering, Madhav been proposed to create secure, efficient, and convenient
Institute of Technology & Science, Gwalior, MP 474005, road transportation system through the vehicular commu-
India nication networks (VCN). With the advancement of wire-
2
Wireless Network and Cloud Computing (WNCC) Lab, less technologies, several improved devices, i.e., advanced
Electronics and Computer Discipline, DPT, Indian Institute sensors and the devices used in Internet of things (IoTs)
of Technology Roorkee, Roorkee, Uttarakhand 247667, India have been incorporated in the vehicular communication
3
School of Computing Science and Engineering, VIT Bhopal system. These advancements address different hazardous
University, Bhopal, MP 466114, India problems associated with the network of intelligent vehi-
4
Department of Computer Science and Engineering, Netaji cles. In the vehicular networks, inter-communication
Subhas University of Technology, Delhi, India

123
2552 Wireless Networks (2023) 29:2551–2575

primarily depends on the cooperation among the vehicles lack the reliability in the transmitted data. In contrast, the
and it is managed using the wireless communication rout- data-based trust models primarily consider the reliability of
ing technology [1]. The routing mechanism usually follows transmitted data to ensure communication network’s
‘‘store-and-forward’’ approach, which may intricate various security [8, 9]. These methods compare exchanged data
network problems like transmission delay, excessive data with the set of messages transmitted by the trustworthy
overhead, and network partition [2]. Data transmission in a nodes based on the source credibility and content similarity
VCN requires special effort as every vehicle should par- to evaluate the reliability of this data. However, this pro-
ticipate in the data transmission process within the avail- cess may lead to an additional delay in the communication,
able scare resources such as network bandwidth, storage, and also increases the cost when a large amount of infor-
and computation power. However, in this process, a few mation has to be communicated. The data integrity of
rogue nodes may acquire all the critical information amidst network is insufficient to ensure smooth and reliable
null contribution in the communication system and also communication until the corresponding network has
consume most of the valuable network resources [3]. enough trustworthy nodes. For selecting appropriate trust-
Moreover, they discard the acquired data that has to be worthy nodes, the hybrid models are widely preferred trust
forwarded to a legitimated vehicle, which may disrupt the models that isolate the rouge nodes from the network and
usual functions of the VCN. The existence of such type of also discard the false transmitted information [10, 11].
rogue nodes in the VCN encounter severe security issues, Moreover, these models are also prone to the above-men-
e.g., user and data privacy, data integrity, and user tioned network challenges (i.e., transmission delay and
authenticity [4]. Therefore, several trust management control overhead).
frameworks have been widely acknowledged to secure the Due to the several advantages of trust-based models,
VCN [5]. In these frameworks, trust is defined as the they become prominent solutions to establish security in a
degree of belief that a node pursues for the other nodes, complex and dynamic internet of vehicles (IoV) Network.
which is a significant factor to maintain the security in the Currently, several researchers are inclined towards making
VCN. It can be grouped into two categories: (i) local trust, the existing trust based IoV models to be suitable with the
and (ii) global trust. For a particular node, local trust is dynamic environment (i.e., interconnections between the
evaluated based on its direct interactions with another vehicles are transient) [12–15]. In traditional trust man-
vicinal node, whereas the global trust is evaluated based on agement models, the trust evaluation methods can be
the other vicinal nodes’ opinion for this node. grouped into two classes: (i) direct trust evaluation (DTE)
In the literature, several methods have been introduced methods, and (ii) cooperative recommendation trust eval-
to protect the VCN that can be broadly grouped into two uation (CRTE) methods. The DTE methods evaluate the
categories: (i) Cryptography based methods, and (ii) Trust trust based on the signal assessment of the communicating
based methods. Cryptography based methods use various vehicles; however, they can be susceptible to a false
existing cryptographic algorithms, e.g., symmetric-key deduction of vehicles’ trust score. On the other hand, the
cryptography, HASH functions, digital signature, and CRTE methods use the mutual cooperation among the
public-key infrastructure (PKI) to ensure privacy, non-re- vehicles to evaluate their trust scores [16]. Further, a
pudiation, authentication, and data integrity. On the other vehicle with maximum trust score is selected as the next
hand, trust-based models utilize the trust value and repu- relay vehicle for transmitting the data. For deploying such
tation of different nodes to build a secure communication scenario, the existing trust management methods employ
network, which provides a secure traffic to enhance the certain structural assumptions that include several intelli-
efficiency of the VCN. Though the cryptographic methods gent vehicles equipped with smart sensors, global posi-
are suitable to protect the VCN network from malicious tioning system (GPS), and navigation system. The
activities such as identity theft, eavesdropping, and data information associated with a vehicle, e.g., directional
tampering, they increase network delay and power con- movement, speed, and location can be easily gathered in
sumption. These limitations restrict their deployment in a real-time through the GPS. Every vehicle with a unique ID
dynamic network with limited resources. However, such can capture the recent position of a target vehicular node
limitations can be efficiently handled through the various through the dedicated location service. Here, a particular
trust-based methods. Based on the trust formulation target, vehicle acquires the unique IDs of its vicinal vehicles in
trust-based models can be grouped in to three subclasses: real-time. However, for security preservation, these IDs are
(i) entity-based, (ii) data-based, and (iii) hybrid trust assigned and advertised periodically for maintaining the
models. Entity-based trust models estimate the node’s trust information up-to-date. This information can be utilized by
level and isolate the low trust value nodes from the network a vehicle for determining the existence of its adjacent
[6, 7]. These methods assume that the trustworthy node vehicles (Fig. 1).
always transmits the reliable information; however, they

123
Wireless Networks (2023) 29:2551–2575 2553

adopt a similarity and consistency-based recommendation

filtering algorithm in the proposed scheme that first detects
the false recommendations from a particular vehicle(s) and
then it discards them, which significantly improves its
performance under the existence of rogue vehicles.
Therefore, the proposed scheme is well suitable for main-
taining the reliability of an IoV network, even under the
presence of rogue vehicles. Moreover, Due to the light-
weight model, it efficiently handles the primary challenges,
e.g., appropriately working under dynamic topology and
efficient use of scare resources. The significant contribu-
tions of the proposed work are as follows.
• Proposed ReTrust Scheme evaluates the direct and
indirect trusts for the vehicular nodes using multi-
dimensional parameters that enhance the accuracy of
the trust evaluation model. It computes the trust of the
vehicular nodes based on their performance concerning
the transmission history and the feedback report from
Fig. 1 Architecture of dynamic vehicular communication networks adjacent vehicular nodes.
• Developed a correlation and similarity measure-based
In a distributed and self-structured VCN, rogue vehicles mechanism that filters the false recommendations for
severely influence the safety of physical IoV devices. Due restraining the malicious behavior of the rogue vehicles
to the various factors (i.e., no centralized control and and also stimulates the vehicles for taking cooperative
infrastructure unavailability), it encounters many chal- actions.
lenges in the deployment of a reliable security mechanism. • Extensive experiments have been performed using the
Though, several solutions exist to maintain secure com- network simulator NS-2 to explore the security aspect
munication in the IoV networks, they fail to establish a and effectiveness of the proposed scheme based on
balanced trade-off between the efficiency and security in various performance parameters.
the network. Therefore, to overcome this problem, in this
paper, we present a lightweight and hybrid trust estimation The rest of the paper is organized as follows. A detailed
method, namely reliability and recommendation trust based description of related work is presented in Sect. 2. The
(ReTrust) security scheme that effectively recognizes the preliminaries, e.g., network model and attack model are
received information’s acceptability for a particular vehi- described in Sect. 3. Section 4 presents the working of the
cle. It formulates a combined evaluation of the direct and proposed ReTrust method. The related parameter settings
indirect trusts in a unified trust model for all the vehicles in and experimental results with their performance analysis
the network. Here, the direct trust evaluation is carried out are presented in Sect. 5. Finally, Sect. 6 concludes the
using the multi-dimensional trust parameters (i.e., trans- proposed work with its future research directions.
mission behavior reliability, communication frequency,
and data reliability), that improves the trust estimation
accuracy. In contrast, the indirect trust is established based 2 Related work
on the recommendation received from the other neighbor-
ing nodes. Recommendation trust is a special category of In the VCN, the connection sessions usually persist for a
direct trust that is received from the vehicles (recom- very short interval of time due to the dynamical changes in
mender) present in the vicinity of the observer node. Since the network topology that makes it challenging to form
these recommendations obtained from the other vehicles appropriate trust relationships between different IoV net-
are not always trustworthy, an effective method is required work’s vehicles. The proposed method follows trust-based
that can distinguish between the false and true recom- mechanism to secure underlying network, therefore, some
mendations. However, considering the aforementioned of the seminal trust estimation-based methods are descri-
trust-based models that exploit the sole direct or indirect bed in the following. As this work relied on a trust-based
trust scheme only rather than including an effective false mechanism, which secures the underlying VCN, here, we
recommendations filtration mechanism with it that may present a few most competent existing works related to
help to improve their performance. Based on this fact, we estimating the vehicles’ trustworthiness and the reliability
of the communication system in the IoV network.

123
2554 Wireless Networks (2023) 29:2551–2575

Generally, VCN’s are infrastructure-less and self-orga- distributed IoV networks. However, the presence of mali-
nized networks that use open wireless communication cious vehicles significantly reduces the security of a self-
technology. These characteristics, along with the use of organized, distributed dynamic network. Indeed, the chal-
open wireless technology in such networks, make them lenges of dynamic network can be better addressed by
dynamical in nature and susceptible to several security employing a self-adaptive security mechanism. Li et al.
issues, e.g., introduction of rogue node, malicious activi- [23] proposed a self-adaptive method, namely attack-re-
ties. To address the malicious activities of rogue vehicles, sistant trust (ART) that identifies and isolates the rogue
Yao et al. [17] proposed an entity-centric trust model that vehicles through estimating the trust of data and the vehi-
employs direct and recommendation-based trust estimation cles in the network. Though ART method shows good
to assess the reputation of a vehicle. It evaluates the trust resistance against the malicious attacks, it may introduce
score for vehicular entities through the weight factor communication overhead in the network. Here, the pre-
associated with the data and vehicle. However, this model dicted trust of vehicle x for vehicle z is expressed as
does not consider the influence of false recommendations. follows.
Soleymani et al. [18] presented a message-based trust P  
y2Sx cos ðx; yÞ  Ry;z  Ry
evaluation model that relied on fog computing concepts. Txz ¼ Rx þ P ð2Þ
Moreover, it incorporates a fuzzy logic scheme to classify y2Sx j cosðx; yÞj

the legitimate message in the VCN appropriately. How- Here, Rx and Ry represent the mean trust scores for
ever, this model is not suitable for dynamic network nodes x and y, respectively.
topology where the network conditions are frequently Recently, Fan et al. [24] introduced a trust-based
changes. Xia et al. [19] presented a trust estimation model framework by employing a certain-factor model to evaluate
that is suitable with dynamic network topology. It calcu- the direct and indirect reputation of the vehicles. It also
lates the trustworthiness of the vehicles based on their incorporates K-means algorithm to identify the legitimate
historical activities that include monitoring the data for- information in the network. Here, the direct reputation of a
warding ratio of an underlying vehicle. The data forward- vehicle is estimated by monitoring its communication
ing ratio is defined as. performance, and is evaluated as follows.
FRðtÞ ¼ Ncor ðtÞ=Nall ðtÞ ð1Þ  
DBx;y ¼ ðax;y  nx;y  ax;y Þ=nx;y ð3Þ
Here, Ncor ðtÞ denotes the cumulative number for correct Here, ax;y denotes the total cooperative communication
packets and Nall ðtÞ represents the total number of packets events and nx;y is stand for the total communication events
originally requested. This model can identify the rouge
between vehicle x and y. And the indirect reputation of a
vehicles and resist them to influence the networks’ per-
vehicle is obtained through the collective reports from the
formance. Though it has improved performance in terms of
neighboring vehicles and road side unites, and is evaluated
transmission delay and data delivery rate, its implementa-
as follows.
tion becomes challenging in the absence of centralized    Pn .
control, limited resources, and dynamic network design. IBx;y ¼ p  q=p þ q  i¼1 ri;y n ð4Þ
Wei et al. [20] proposed a trust-based framework that
significantly improves networks’ throughput, communica- Here, p and q denote the number of vehicles with pos-
tion delay, delivery rate, and minimizes the network con- itive reputation and number of vehicles with 0 or negative
trol overhead in the data transmission. Shankaran et al. [21] reputation, respectively.
proposed a trust-oriented framework for the mobile ad-hoc Kerrache et al. [25] presented a new architecture,
networks (MANET). Nevertheless, it is not suitable to namely T-VNets, to secure the vehicular ad-hoc networks
deploy in the IoV networks that accommodate time-vary- using a trust-based model. T-VNets estimates the traffic
ing topology and high-speed moving vehicles. Moreover, it density, trust of entities, and distribution of dishonest nodes
requires a proper stable connection to process the infor- in the VCN to eliminate the dishonest nodes from the
mation regarding the identity verification; however, in vehicular network.
reality, high speed of moving vehicles and the dynamic In recent years, entity-based trust evaluation methods
topology makes it difficult to process such identity verifi- become more popular to implement the trust management
cation information effectively. To deal with the afore- framework in a dynamic IoV Network. Depending upon
mentioned issues, symmetric encryption-based schemes are the requirements, sometimes the network deploys all the
utilized for providing secure identity verification (authen- nodes to evaluate the trust scores; sometimes only a set of
tication) in the network. Chuang et al. [22] presented a selected nodes (also known as observer/verifier nodes) are
distributed trust-based authentication mechanism which deployed to perform the same task. Sedjelmaci et al. [26]
appropriately provides the privacy and security in the proposed an entity based, lightweight intrusion detection

123
Wireless Networks (2023) 29:2551–2575 2555

scheme for the VCN. It uses the idea that the observer node recommendation trust-based defence scheme to protect the
monitors the neighboring nodes behavior within its defined network from the attacks that propagate false recommen-
zone. Here, a set of nodes are deployed in the network to dations in the network. It effectively deals with the rogue
observe the communication behavior and exchange their nodes that involve in such attacks to propagate the false
reports to work in cooperative manner. Though it works recommendations for disrupting the network’s functions.
effectively where the networks are cooperative and service Sharef et al. [31] proposed a routing mechanism that is
oriented. However, it may fail to deliver the desired results responsible for selecting a trustworthy and robust mobile
when any of the observer nodes got compromised. Khan gateway. It uses the characteristics of node’s mobility and
et al. [27] proposed a method to distinguish selfish vehicles different routing factors to ensure secure communication in
from the VCN based on their trust scores. Firstly, a vehicle an infrastructure-oriented network. For securing a VCN,
is elected as a verifier node that monitors the vicinal node’s most of the methods discussed above implement it by
communication behavior. The election of verifier node is including the trust management and authentication proto-
carried out using the parameters such as load (LD ), distrust cols. The Cooperative environment is suitable to imple-
score (Dv ), and distance (Ds ) and it is expressed as: ment the security framework in the dynamic network like
Dp ¼ w1 LD þ w2 Dv þ w3 Ds , where w1 , w2 , and w3 are the VANETs [32, 33], but most of the models ignored the
weighted coefficients of LD , Dv , and Ds , respectively. vehicle cooperation factor [34]. The combination of direct
Secondly, the elected verifier node estimates vicinal node’s and indirect reputation is a standard approach to build the
distrust level through the multivalued parameters and the trust model to handle the rogue nodes in the network
estimated distrust values are forwarded to a specific [35, 36]. Table 1 summarizes the various existing security
responsible authority that isolates the nodes having high methods based on various parameters.
distrust values, i.e., rouge node. However, this model High mobility of vehicles may produce frequent dis-
increases the communication overhead due to the transfer connection in the communication links that causes long
of trust reports. delays and a low delivery rate with various security vul-
Primarily there are two variations of trust management nerabilities. Due to these challenges, design of an optimal
models in the vehicular networks. The first variation is secure communication model that can reliably transmit the
based on the evaluation of direct trust among the nodes sensitive data is still a critical issue. In IoV networks, it is
referred as direct trust model. Here, a particular node still an open concern to identify a reliable data forwarding
evaluates the trust score of another node that is present in mechanism that can reduce the influence of rogue nodes. In
its direct communication range. Gharib et al. [28] presented most of the aforementioned trust-based security models
a trust model that analyses the trust in cyber physical [17, 27, 31, 32], the evaluation of trust solely depends on
system (CPS) using structured assurance cases based on the the single parameter. However, single parameter-based
attributes such as claims, arguments, and evidence. It estimation schemes face problem to accurately reflect the
improves the network’s precision by performing the time- vehicles’ performance under the trust evaluation.
series analysis for dynamically tuning the weight constants The existing trust-based models [27–29] lack the coop-
associated with the assessment parameters. However, it did eration among the vehicles and ignore the selfish behaviour
not consider the significance of inner relationship between of vehicles. However, both these factors are desirable to
several attributes on CPS. The direct trust-based model take into consideration for implementing the security in a
significantly improves the routing performance due to the trust-based model. In VCN, the combination of direct and
low control overhead in the varying network. Yan et al. indirect trust is a promising approach in the formation of
[29] proposed a trust-based model to improve the effi- trust model to resist the selfish behaviour of the vehicles.
ciency of a routing protocol in the VCN. It evaluates the As discussed earlier, indirect trust estimation is based on
trust by considering the sole direct trust parameters via the recommendation received from neighbouring vehicles.
comparative analysis, and neglects the use of recommen- However, it requires an efficient mechanism to filter out the
dation trust. The second variation considers both the direct false recommendations that should be incorporated when
trust and the recommendation of trust reports from the designing a trust-based security scheme for VCN. A trust
vicinal vehicular nodes to estimate the effective trust of the estimation model should be efficient and accurate so that
respective vehicle. Here, a particular node considers the the desired communication could established in an under-
opinion of the other nodes present in its vicinity. However, lying network. Further, in the presence of the rogue nodes,
this type of trust models usually suffers from false rec- it should be able to improve the critical networks’ param-
ommendation attacks. In recent time, several methods have eters such as delivery rate, transmission delay with low
been introduced to handle the false recommendation resource consumption.
reports in the VCN. Shabut et al. [30] proposed a The proposed ReTrust scheme considers multiple trust
parameters to upgrade the security of the network. In order

123
2556 Wireless Networks (2023) 29:2551–2575

Table 1 Summary of the existing security methods

Author(s) Security model Performance criteria Simulation tool Attacks Entity/ Remarks
Data-
centric

Hu et al. Reliable Trust Trust Score, Reputation MATLAB Badmouth/ Entity Feedback based trust
[3] recommendation Value, Detection rate Ballot-stuffing Centric algorithm not ensures the
framework attacks, On–off privacy of data
attacks
Huang Trust model based End-to-End Delay, NCTUns Badmouth attack Entity Model may introduce some
et al. [7] on social network Anomaly Ratio (AR), Simulator based delay in decision making
approach False Positive Rate
(FPR)
Wan et al. Events reports Malicious node rate VEINS Sybil attack, Data- Only data trust has been
[8] based Trust model Badmouth Centric considered in algorithm
attack
Gazdar Distributed trust- Expected State (ES), NS-2 Eavesdropping Hybrid No use of recommendation
et al. [11] based protocol Probability (PS) and fake events Model values
Yao et al. Dynamic entity- Packet delivery ratio, Java Language Blackhole Entity Recommendation trust
[17] centered, Average path length, on Window 7, attack, Centric filtering scheme not used
weighted trust Average E2E delay Vanet- Selective
model MobiSim forwarding
attack
Soleymani Fuzzy based trust Selection of NS-2 with Bogus Message, Entity Fog nodes have been
et al.[18] model in VANET Trustworthy nodes SUMO & Message Centric deployed to assess the level
with Fog (%) MOVE Alteration, of accuracy of the event’s
Computing Obstacles location
Khan et al. Entity-centered Delay, Throughput, Network Packet dropping/ Entity No use of recommendation
[27] Trust Packet delivery rate Simulator-2 duplication Centric trust
Management attack
Model
Yan et al. Lightweight Average acceptation OMNet Bad mouth Entity The fully self-organized trust
[29] Statistical trust rate, Average Trust attack trusts model effective with certain
model Value with attacks
data
trust
Baiad et al. Cross-layer model Detection rate, False MATLAB Black Hole Entity In this cross-layer design, no
[33] based on positive rate Attacks Centric consideration of monitoring
cooperation node’s reputation

to enhance the accuracy of direct trust evaluation, we 3 Preliminaries

consider multiple parameters for trust estimation rather
than a single parameter unlike most of the existing trust- In this section, we describe the preliminaries like various
based methods. For evaluating the indirect trust, the pro- roles performed by the vehicles in the VCN, network
posed method considers the recommendation trust; mean- model, and attack (adversary) model that are required to
time, it ensures the cooperation among the several vehicles evaluate the performance of the proposed ReTrust method.
in the VCN effectively. To deal with the rouge vehicles, a
similarity and consistency centred recommendation filter- 3.1 Network model
ing mechanism is introduced that identifies and discards the
respective false recommendations. The filtration of false A vehicular ad-hoc network is generally a collection of
recommendation enhances the accuracy of indirect trust heterogeneous wireless devices, e.g., multiple sensors and
evaluation. Moreover, the ReTrust scheme is lightweight in computing/processing devices deployed in the vehicles.
nature and it is suitable to deploy in dynamic VCN under Such network provides continuous monitoring and sharing
limited resources. The primary emphasis of the work pro- the road conditions in the transportation systems. In order
posed here is to present the data forwarding method at the to realize this phenomenon, we design a network model to
network layer using a trustable communication path. demonstrate the proposed ReTrust scheme under the

123
Wireless Networks (2023) 29:2551–2575 2557

presence of rogue vehicles, where all the networks’ vehi- range of node s is estimated through the following
cles can travel in a defined route. Here, all the considered expression [29]:
nodes are categorized into four groups (i) observer nodes, pR2
(ii) recommender nodes, (iii) object nodes, and (iv) selfish p¼ ð5Þ
xy
or rogue nodes [5].
Let ‘i’ be an evaluating node referred as an observer The relative position stability of two vehicles within
node that wants to monitor the behavior of a node ‘j’ (re- one-hop range can be estimated in terms of relative speed
ferred as an object node) whose trust has to be evaluated Vk and n neighboring nodes using the following relation:
(Fig. 2). In other words, node ‘i’ returns the trust value for Xn
1
node ‘j’. Here, an observer node can be characterized as a Sp ¼ Vk  ð6Þ
k¼1
n
node having higher personal trust value and low mobility
that evaluates the trust for the other nodes in the network. The probability that vehicle s has c vicinal vehicles
Further, to distribute the computation load evenly, a set of (from n-1 vehicles) can be evaluated through the binomial
nodes have been selected as the observer node. In this distribution as:
work, the communication in VCN is based on the multi-  
n1 c
hop network model, which means that any node can Pc ¼ p ð1  pÞn1c ð7Þ
c
directly communicate with adjacent nodes within their
communication range. The message communication Here, c is a random variable and E½c represents the
between two non-neighbor nodes is forwarded with other average number of neighboring vehicles for any particular
intermediate nodes. The intermediate relay nodes not just vehicle in the network.
forward the messages but also can process the information It is assumed that the vehicles are moving according to
based on their own opinion. The evaluation of trust value is the Random Way point model, where link between any two
based on the observer node’s observations and recom- vehicles are formed using the lognormal model given as
mendation from the neighboring nodes. The neighboring follows [37]:
nodes which provide the recommendations referred as   
1 log dði;jÞ
‘‘recommender nodes’’. After the evaluation of the trust pdði;jÞ ¼ 1  erf m ð8Þ
2 e
score for the object node, it will be declared as a trust-
worthy or rogue node [6]. Here, dði;jÞ is the normalized distance between vehicle i
In network model, let us assume that n vehicles are and j and e can be estimated as e = r=g, r and g denote the
distributed randomly in the area x  y square meter. The standard deviation for the signal power fluctuations and
transmission range of each vehicle is R, where x; y  R. path-loss exponent, respectively. Further, m is defined as:
Here source and destination vehicles are represented as s 10
and d, respectively. The probability of node k to be in the v ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð9Þ
2 log2 10
A vehicle is called as the neighbor of another vehicle if
it is located in the transmission range of that vehicle. The
neighboring vehicle can exchange the messages success-
fully to one another. In the simulation study, following
assumptions have been considered:
(i) All the nodes are set into the promiscuous mode so
that they can monitor all the traffic generated in
the network.
(ii) All the nodes are equipped with GPS and other
sensors with DSRC protocol and have the same
memory, communicating and processing
capabilities.
(iii) All the nodes have the same communication range
and their communication ability may be limited
due to the consideration of specific wireless
technique.
(iv) The nodes in VCN have limited energy, low
computation and storage capabilities.
Fig. 2 Network model for VCNs

123
2558 Wireless Networks (2023) 29:2551–2575

3.2 Attack model recommendation. In this work, it is assumed that only

rogue nodes in the network transmit the false
Usually, due to their distributed nature, VCN’s are sus- recommendations.
ceptible to various types of attacks. In these networks,
attacker may be an insider or an outsider. Similar to the
wireless networks, a VCN is also vulnerable to several
4 Proposed ReTrust method
selfish attacks, e.g., Sybil attack, wormhole attack, and
denial of service (DoS) attack. Moreover, it is observed
The secure data dissemination is the foremost concern in
that a trust estimation model in VCN is also susceptible to a
Vehicular communication system to transmit critical data
number of selfish node attacks, i.e., false information attack
(such as an alert or emergency warning message). The
and on–off attacks. The primary aim of an attacker includes
proposed scheme can be used to build a trustable path by
modifying the information, forwarding false recommen-
selecting the best suitable next relay node for data trans-
dations, intercepting the message transmissions, etc. [34].
mission. This will improve the reliability and efficiency of
In order to investigate the performance of the proposed
the communication network.
ReTrust scheme and the existing models under the pres-
Though the existing trust-based models [17–19] incor-
ence of selfish/rouge node in the VCN, we simulate them
porate different security strategies in their work, they still
with two types of selfish/rouge node attacks (i) packet
suffer with various limitations such as high overhead,
drop/modification, and (ii) false recommendation. In
higher resource requirement, and low prediction accuracy
packet drop attack, a rogue node can intentionally drop the
due to the consideration of single parameters, respectively.
packets or can alter the data to generate false information.
Here, real-time prediction accuracy plays a vital role to
It introduces several problems in the network, e.g., the
justify the performance of an underlying method. In order
complete loss of a critical message, propagation of false
to achieve the desired trust prediction accuracy, this work
information that may cause serious performance degrada-
proposes a secure trust framework, namely reliability and
tions concerning to the networks’ functions. In contrast, in
recommendation trust based (ReTrust) method that
false recommendation attack, a rogue node intentionally
employs vehicle cooperation-based trust evaluation and
forwards bad recommendations about the object node in
recommendation filtrations. It is primarily consisting of
the network. In this scenario recommendations from rogue
four modules: (i) direct trust (DT) calculation, (ii) simi-
nodes cannot reflect the genuine opinions against the object
larity and consistency evaluation for recommendation trust
node. More description about the attacks covered in this
(RT) evaluation, (iii) indirect trust evaluation, and (iv)
work is given below:
combined trust (CT) or effective node trust estimation. The
a. Packet drop/modification attack: Rogue nodes can working procedure of the proposed ReTrust method is
drop or alter the communication packets before shown in Fig. 3.
reaching to the destination node. The source node Here, to establish a secure and trustable environment,
should verify the trustworthiness of the immediate these four modules are interacted with each other to
neighbors before sending the data to any intermediate mutually exploit their corresponding calculated trust
node. Rogue vehicles can carry out a variety of packet information. In the first module, i.e., direct trust module
drop attacks, including the Blackhole, Grey-hole, and evaluates the communication reliability using the trust
Wormhole attacks. values of all the corresponding active relay nodes (data
b. Bad mouthing attack/Recommendation Attack: In a forwarding nodes between the source and destination) in
bad-mouthing attack, rogue nodes can send the wrong the network. In the second module, i.e., the recommenda-
recommendation trust value to the other nodes. Assume tion trust module performs the evaluation of recommen-
that the recommendation trust value is communicated dation trust through the various recommendations collected
in binary. In that situation, the attacker node can from the present neighbouring vehicles. In the third mod-
simply flip the one-bit recommendation information to ule, i.e., the CDM module identifies the true recommen-
tamper with the honest node’s reputation, causing trust dation using the proposed consistency and similarity
evaluation to be inaccurate. Furthermore, rogue nodes measures among all the recommendation values collected
might employ this form of attack to improve their from various neighbours. In the final module, it evaluates
reputation or influence in the network, which is the combined trust or effective trust of the respective node
referred to as a bad mouth attack. A rogue node, on and then decides the trustworthiness for this node. Finally,
the other hand, can provide other selfish nodes a better it broadcasts the status of the respective node in terms of

123
Wireless Networks (2023) 29:2551–2575 2559

Fig. 3 Trust Framework for proposed algorithm

binary values, where 0 and 1 represent that the node is not Therefore, the proposed framework evaluates the direct
trustworthy and trustworthy, respectively. Here, instead of trust by incorporating three parameters include (i) trans-
large information, the utilization of binary values helps to mission reliability, (ii) communication frequency, and (iii)
reduce the overhead in the network. Table 2 describes the content reliability. The computation of these parameters is
various mathematical symbols used to describe the math- comparatively feasible in a real-time scenario and also
ematical model of the proposed algorithm. The detailed suitable to obtain the desired trust accuracy.
description of each module of the ReTrust scheme with
their formulation is given below. 4.1.1 Transmission reliability

4.1 Direct trust The data forwarding behaviour of any node is used to
determine its transmission reliability. Here, the first con-
Most of the existing trust evaluation methods [31, 32] sidered parameter, i.e., transmission reliability of a relay
incorporate single or limited communication parameters node is the observer node’s (X) own experience concerning
for the sake of lower communication overhead and trust to the respective node (y) and is evaluated by computing
computation complexity. However, the restricted number the data forwarding performance of node ‘y’. Here, node
of these parameters leads to limit the performance of the ‘X’ monitors the data traffic of node ‘y’. The transmission
corresponding methods in terms of desired trust accuracy. or communication reliability of node ‘y’ is estimated based

123
2560 Wireless Networks (2023) 29:2551–2575

Table 2 Description of various symbols used

Symbols Description

sðX;yÞ Transmission Reliability of node ‘y’ evaluated by node ‘X’

TðdX;yÞ Direct trust of node ‘y’ calculated by node ‘X’
TðrX;yÞ Recommendation trust of node ‘y’ calculated by node ‘X’
TðcX;yÞ Combined trust of node ‘y’ calculated by node ‘X’
RfðX;yÞ Recommendation factor

BTðdX;yÞ Binary Value of Direct trust

BTðrX;yÞ Binary Value of Recommendation trust
BTðcX;yÞ Binary Value of Combined trust
Tðrk;yÞ Trust of node y calculated by node k
Thd ; Thr ; Thc Threshold for direct, recommended and combined trust
Pl Priority levels of vehicles
Wv Weight of the vehicle
pf andd
SR ðM; Zk Þ Similarity measure between the recommendation of node M and Zk
DR (M, Zk) Dissimilarity measure between the recommendation of node M and Zk
corrðM;Z Þ Correlation computation between the recommendation of node M and Z
Ryv List of recommendations (Recommendation vector) for vehicle y at time t
Ary The average recommendation trust vector
corrðAr ;T r Þ Measure of correlation between single node recommendations and average recommendations vector
y k

Assume the following scenario: Node ‘X’ wants to assess the Effective trust value of node ‘y’ based on direct contacts with node ‘y’ and
feedback from other nodes. Here, node ‘X’ is an observer node deployed to determine the trustworthiness of object node ‘y’, while other nodes
called recommender nodes (e.g., Node ‘k’) assist the observer node by sending recommendations

on the analysis of the communication traffic generated number of packets forwarded without an alteration (called
through node ‘y’. Let us assume that node ‘y’ sends wt (y) valid packets). After every direct interaction between node
correct packets (some packets are altered, dropped, or by ‘X’ and node ‘y’, both the TðX;yÞ and MðX;yÞ are updated
node y) out of total zt (y) packets at the time’t’. The using the following expressions.
expression to evaluate the data forwarding rate is as 0
TðX;yÞ ¼ TðX;yÞ  e#Dt þ Eðt X;yÞ ð12Þ
follows.
0
 
W t ð yÞ MðX;yÞ ¼ MðX;yÞ  e#Dt þ 1  Eðt X;yÞ ð13Þ
vðX;yÞ ¼ ð10Þ
Zt ð yÞ
Here, e#Dt represents the exponential decay with 0 as the
Now, the transmission reliability for a node ‘y’ can be decay factor, and Dt denotes the interval for a trust update.
evaluated by expanding Eq. (10), here the direct interac- The reputation score is represented by Eðt X;yÞ having a
tions between nodes ‘X’ and ‘y’ is represented using the
binary value relay upon the direct interactions. Values ‘0’
following expression that is based on the beta probability
and ‘1’ represent the unsatisfied reputation and the satisfied
distribution:
reputation, respectively. During the periodic monitoring,
TðX;yÞ the observer node updates (increased or decreased) the
sðX;yÞ ¼ ð11Þ
TðX;yÞ þ MðX;yÞ penalty/reward to the reliability value of the node over time
based on its performance.
Here, MðX;yÞ denotes the no. of packets either altered or
dropped (called malicious packets), and TðX;yÞ is the

123
Wireless Networks (2023) 29:2551–2575 2561

4.1.2 Communication frequency TðdX;yÞ ¼ w1 sðX;yÞ þ w2 uðX;yÞ þ w3 dðX;yÞ ð17Þ

After the evaluation of transmission reliability, we estimate Here, w1 , w2 , and w3 are the weight factors of the
the communication frequency of a particular node. Com- respective parameters that are used to assign the priority
munication frequency is the second parameter considered level to each of these parameters at the time of trust
for the evaluation of direct trust that represents the degree evaluation.
of interaction between two nodes. If communication fre- In this work, to decrease the communication overhead,
quency for a particular node is higher than the other we convert the trust score into binary values so that only 1
adjacent nodes, its probability of forwarding the data is bit is required to send the recommendation trust informa-
also higher. The communication frequency is defined as: tion along with the respective packets. Here, a node
uðX;yÞ ¼ kðX;yÞ =x ð14Þ requires maximum (n-1) bit with the message for the rec-
ommendation trust data, where ‘n’ is the number of current
Here, uðX;yÞ denotes the communication frequency of neighboring nodes. When any observer node wants the
node ‘y’, kðX;yÞ and x represent the number of transmis- recommendation values from other neighbors, they forward
sions with node ‘X’ and with other nodes, respectively. the original recommendation trust value to the observer
node. In all the other cases, they convey the binary value.
4.1.3 Data reliability The conversion of the direct trust value into binary value is
carried out by comparing the trust value with a predefined
Now, we evaluate the third considered parameter, i.e., data threshold value (Thd ) and is given by. If TðdX;yÞ  Thd then
reliability that is directly related to the trust of the data BTðdX;yÞ ¼ 0 (node ‘y’ may be a rogue node), Otherwise
originated/forwarded from a particular node. In the pro-
BTðdX;yÞ ¼ 1.
posed method, to estimate the trustworthiness of a node, we
monitor the consistency of data originated from it. Gen-
erally, different data originated from the same source have 4.2 Recommendations trust
some correlation. Moreover, the data transmitted between
two neighboring nodes always have the ‘‘spatial correla- Each node in the vehicular network sends the recommen-
tion’’ in the same region. Usually, the content values in the dation value of their immediate neighbors on request, to the
transmitted information follow some statistical distribution, observer node. For the efficient functioning of the VCN,
i.e., normal distribution. Therefore, in this method, we we require to distinguish the rogue vehicles before the
assume that the data in the corresponding messages follow indirect trust evaluation. A rogue vehicle may forward false
the normal distribution. The probability density function recommendations, to disturb the network functioning or to
(PDF) for these data sets can be defined using the increase its reputation in the network. However, this type
expression of normal distribution given as: of selfish behavior can be detected by analyzing every
recommendation obtained from the recommender vehicles.
1 ðklÞ2
f ðkÞ ¼ pffiffiffiffiffiffi e 2r2 ð15Þ The recommendation trust values transmitted by rogue
r 2p nodes always have a large difference compared to the
Here, k denotes the attribute value for cd content data honest nodes.
item, mean l represents the most representative value of After receiving the recommendation from every node,
data that reflects the similarity value of data and is sup- the node compares each recommendation with the average
posed to have highest trust value. If the data value is also recommendation trust for any particular node. This rec-
closer to the mean value (l) then the trust of the data can be ommendation filtering scheme enchases the accuracy of
considered relatively higher side. Therefore, we can define trust evaluation. Various methods have been developed for
the data reliability for data items using the following similarity mining. In this regard, few methods use the
expression [37]. collaborative filtering approach [23] to find the similarity
cd between non-linear data, whereas, some of them adopted
dðX;yÞ ¼ 1  2 r f ðkÞdk ð16Þ the Euclidian based method [38]. Moreover, to filter the
l
false recommendation trust values, Liang et al. [36] sug-
Finally, using the evaluated parameters, i.e., transmis- gested the K-Nearest Neighbour (KNN) algorithm. The
sion reliability, communication frequency, and data relia- existing recommendation filtration schemes either use the
bility, we evaluate the direct trust for the given node using similarity or the consistency measure at a time. In contrast,
the following expression. this work categorizes the false recommendations based on

123
2562 Wireless Networks (2023) 29:2551–2575

the similarity and consistency measures because of sim- the message. The observer node X calculated the similarity
plicity and broad applicability. The proposed recommen- between received recommendations and self-opinion (if
dation filtration scheme uses the similarity measure to self-opinion has been formed). Otherwise, observer node X
check the similarity for recommendation and the consis- evaluates the similarity between the recommendation of
tency measure to validate the consistency of the recom- node k and the recommendation received from other nodes
mendations from the particular recommender node. presented in the locality of y. This method identifies the
In the VCN, the vehicle nodes exhibit diverse and false recommender from the set of recommender nodes.
complex security behaviors. The evaluation of a vehicle’s Vehicle X receives a set of recommendations from
security behavior is typically based on characteristics such vicinal nodes for the object vehicle y (Fig. 4). Node k sends
as information security, safety driving, information the recommendation to node X about the various other
authenticity and accuracy, and so on. The recommendation nodes presents in the vicinity. Other nodes such as Z1,
vehicle assesses the object node depending on each secu- Z2…. Zn also send the recommendation values to the
rity behavior. Furthermore, the evaluated score is stored observer node X. All the nodes send their recommendation
and forwarded as vector data, referred to as a recommen- value within the message with one extra field.
dation vector. Usually, recommender vehicles can be Here, it is supposed that node ‘X’ has not formed its
classified into two categories: honest recommendation own opinion about node ‘y’. Node X maintains the list of
vehicles and selfish recommendation vehicles. The rec- neighbors and filters the mutual nodes that are also closer
ommendations received from the rogue vehicles are mainly to node ‘y’ denoted by N = {Z1, Z2…… Zn}. After that,
responsible for the inconsistency in the evaluation process. node ‘X’ calculates the similarity between the set of rec-
We present a selfish recommendation vehicle identification ommendation values received from node ‘k’ (i.e., Rk) and
approach based on identifying false recommendations to the collection of recommendations received from mutual
eliminate the influence of selfish recommendations in neighbors set {Z1, Z2…… Zn} denoted by RZ1,
VCN. This algorithm has two phases: We check the com- RZ2,…….RZn. The similarity between k and Zi can be
parability of the received recommendations from the par- expressed as [39]:
ticular vehicle with the recommendations from other jRk \ RZi j jRk \ RZi j
vehicles in the first phase. We compare filtered recom- SR ðk; Zi Þ ¼ ¼ ð18Þ
jRk [ RZi j jRk j þ jRZi j  jRk \ RZi j
mendations to average recommendations using a correla-
tion coefficient in the second phase. The recommendations To find the dissimilarity between the recommendations
with the larger deviations are considered bogus recom- from k and Zi, we can use the following equation:
mendations and are discarded. The following subsections jRk [ RZi j  jRk \ RZi j
provide a more in-depth look at the topic. DR ðk; Zi Þ ¼ 1  SR ¼ ð19Þ
jRk [ RZi j

4.2.1 Recommendation similarity We denoted the average similarity score of recom-

mender node ‘k’ with other ‘n’ neighbours at the time ‘t’:
Pn R
The similarity score of the recommendations obtained from   S ðk; Zi Þ
the vicinal nodes, as well as the consistency of the sender’s AVG SRt ðkÞ ¼ i¼1 ð20Þ
n
recommendations, are used to evaluate recommendation
If Node ‘X’ has evaluated its recommendation about
trust. Once a node has formed an opinion about its neigh-
object node, then we can also directly compare the simi-
bours, it broadcasts that opinion in binary form along with
larity between the recommendations of ‘k’ and ‘X’ as
SR ðk; X Þ. The similarity score for node k, i.e., SRt ðkÞ,
describes the recommendation reliability of the node at the
given time instant. The overall similarity score for node k
till time t is evaluated using:
nRt ðkÞ ¼ hSRt ðkÞ þ ð1  hÞnRt1 ðkÞ ð21Þ
Here, nRt ðkÞ: defines the similarity score for node k at
time t. The similarity score provides a sufficient measure to
estimate the trustworthiness of recommendations. More-
over, we estimate the consistency of node k to identify the
true recommendations.

Fig. 4 Recommendations from neighbors

123
Wireless Networks (2023) 29:2551–2575 2563

4.2.2 Recommendation consistency related. The procedure to calculate the correlation between
two random values (X and Y) is given below [40]:
We suggested a correlation-based false recommendation Pn   
i¼1 Xi  X Yi  Y
detection method to filter out the false recommendations corrðX;Y Þ ¼ qðX;Y Þ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn  2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn  2ffi
from all of the recommendations acquired from the vicinal i¼1 X i  X i¼1 i Y  Y
nodes. Firstly, it calculates the average trust recommen- ð26Þ
dation parameter by utilizing the mean score value of each
obtained trust recommendation from different sources. covðX; Y Þ
corrðX;Y Þ ¼ qðX;Y Þ ¼ ð27Þ
After receiving the recommendation from every node, it SX S Y
compare each recommendation for any particular node We have calculated the correlation coefficients between
with the average recommendation trust. The proposed average recommendation trust Ary and single-vehicle node
approach has recognized fake or selfish recommendations
recommendation trust Tkr . The evaluation of the correlation
in the network by comparing the recommendation values.
coefficient can be done using:
List of recommendations (Recommendation vector) for        
vehicle y at time t can be represented by: cov Ary ; Tkr ¼ E Ary ; Tkr  E Ary E Tkr ð28Þ
Ryv ¼ RðZ1 ;yÞ ðtÞ; RðZ2 ;yÞ ðtÞ; RðZ3 ;yÞ ðtÞ; RðZ4 ;yÞ ðtÞ. . .. . .. . .. . .  
cov Ary ; Tkr
ð22Þ corrðAr ;T r Þ ¼ qðk;Zi Þ ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð29Þ
y k  qffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
Here, Ryv represents recommendations trust vector from D Ary D Tkr
vehicle Zi for vehicle y at time t, and RðZ1 ;yÞ ðtÞ denotes
recommendation value component obtained from vehicle Here, corrðAr ;T r Þ : Measure of correlation between single
y k
Z1 to vehicle y at time t. The trust score for a certain node recommendations and average recommendations
vehicle may change over time. A specific vehicle may vector, and SX & SY : denotes standard deviations of the
transmit different recommendation scores for the same given sample.
object vehicle at different periods. The observer node To identify the false recommendations transmitted by
continuously receives several recommendations from dif- selfish nodes, we compare the evaluated correlation coef-
ferent vehicles as well as multiple recommendations from ficient with the threshold value. The threshold value has
the single vehicle over time regarding a particular object been decided to divide the range of jqðk;Zi Þ j into two
vehicle to evaluate the trust. Various recommendations for intervals. Based on the threshold values, we determine the
the vehicle y from a specific vehicle k at a different period trustworthiness of recommendations. Let us assume that
can be grouped into one list in the given format: the threshold value for coefficient is Dq, and two intervals
Tkr ¼ Rðk;yÞ ðt1 Þ; Rðk;yÞ ðt2 Þ; Rðk;yÞ ðt3 Þ; Rðk;yÞ ðt4 Þ. . .. . .. . .. . . can be taken as (0 to Dq) and (Dq to 1). The value of Dq is
decided based on the experiment analysis and adjusted
ð23Þ according to the scenario. Based on the below two cases we
The average of all recommendation from vehicle k can identify the selfish vehicle sending the false
based on the r number of recommendations received rela- recommendations:
ted to vehicle y can be evaluated using the following • When qðk;Zi Þ belongs to the interval (0 to Dq) that means
relation:
the correlation coefficient of Ary ; Tkr does not show the
1X r
linear relationship. It denotes that the recommendation
gT ¼ RðZi ;yÞ ðtk Þ ð24Þ
r i¼1 trust from vehicle k is not considerable, and vehicle k is
declared as a selfish vehicle.
The average recommendation trust vector can be • When qðk;Zi Þ belongs to the interval (Dq to 1), which
established:
means received recommendation is linearly correlated
Ary ¼ RðZi ;yÞ ðt1 Þ; RðZi ;yÞ ðt2 Þ; RðZi ;yÞ ðt3 Þ; RðZi ;yÞ ðt4 Þ. . .. . .. . .. . . with average recommendations. We can consider the
ð25Þ recommendation as to the true recommendation.

When the observer node receives any recommendation Furthermore, if the correlation coefficient falls in the
score from node k, it uses the correlation coefficient range of (Dq to 1), we can consider it the true recom-
technique to validate the consistency of the received rec- mendation. To further verify it, we can apply a regression-
ommendation. The ‘‘correlation coefficient’’ q is used to analysis based statistical method to find the relationship
measure the relationship between two random variables. between two variables (suppose X and Y). Presuming that
Moreover, it indicates how strongly two variables are the value of X is known and that the value of Y is not

123
2564 Wireless Networks (2023) 29:2551–2575

completely specified, but that the values tend to cluster factor is calculated first, and then the indirect trust value is
around a certain average value. This can be expressed evaluated. In terms of negative and positive suggestions,
mathematically as: f ð xÞ ¼ EðY=X ¼ xÞ. Suppose {xi } and we estimated the recommendation trust factor [42].
{yi } be the sample set for the variable X and Y respec- Pt ð y Þ þ 1
tively. We can describe the linear relationship between x RfðX;yÞ ¼ ð35Þ
Pt ð yÞ þ Nt ð yÞ þ 2
and y using the equation for a straight line as [41]:
y¼sxþc ð30Þ Here, Nt ð yÞ represents the number of negative recom-

Here, s and c denote the slope and the intercept value, mendations from the other neighbouring nodes for node y,
respectively. If we consider the error deviation named Pt ð yÞ denotes the number of positive recommendations for
residues (ei ), then we have the relation represented as: node y.
Further this scheme evaluates the indirect trust for the
y i ¼ s  x i þ c þ di ð31Þ
object node based on the obtained trust score, weight of the
We use the least square method for a better approxi- recommender node and recommendation factor evaluated
mation, where sum Sd is represented as squares of error from several recommendations.
deviations. Pn  d

i¼1 W Z  T ðZi ;yÞ
X n X
n
TðrX;yÞ ¼ Pn
i
 RfðX;yÞ ð36Þ
Sd ¼ di2 ¼ ðyi  sxi  cÞ2 ð32Þ T d
i¼1 ðZi ;yÞ
i¼1 i¼1
Here, n is the number of neighbouring nodes from which
The slope ‘s’ can be evaluated using the relation as
given below.
the observer node receives the recommendation trust value.
covðX; Y Þ Furthermore, the weight factor’s multiplication ensures that
s¼ ð33Þ
Var ð X Þ the recommendation value is prioritised based on the priority
of the corresponding vehicle. In vehicular networks, there are
Similarly, here we apply the regression analysis with
many different types of nodes. Some vehicles have special
average recommendation trust and single vehicle recom-
priority in road transport (such as army vehicles, ambulances,
mendation trust.
  and so on), while others behave like regular vehicles (Such as
cov Ary ; Tkr cars, taxi, etc.). As a result, based on the priority rating, we can
s¼   ð34Þ assign some weight to each vehicle. The data transmitted by a
Var Ary vehicle with a higher priority has a high level of reliability.
The various priority levels of the vehicles in the network are
If the evaluated slope of the linear regression is close to defined in this section. The lower the number in this method,
1, the received recommendation is classified as a true the greater the vehicle’s priority, as seen below:
recommendation. However, if the value of slope s is not 8
close to 1, then the received recommendation can be < 1; Vehicles with higher priority
classified as a false recommendation. With the use of Pl ¼ 2; Vehicles with medium priority
:
recommendation similarity and consistency-based tech- 3; Vehicles with lesser priority
nique, Algorithm 1 defined the pseudo-code for estimating In addition, the weight of the node (or vehicle) in the
the object node’s recommendation trust. From the collec-
tion of suggestions received from the neighbour set, it network is determined by the priority level.
determines the genuine recommendation. Moreover, an 8
< 1; WherePl ¼ 1
effective recommendation value has been produced using
Wv ¼ 0:5; WherePl ¼ 2 ð37Þ
all of the legitimate suggestions for the object node. :
0:2; WherePl ¼ 3

4.3 Indirect trust For further calculation, we convert the value of indirect

The estimation of the indirect trust of the vehicle node is trust into a Boolean value: If TðrX;yÞ  Thr then BTðrX;yÞ ¼ 0,
based on the recommendation trust received from other (node ‘b’ may be malicious node), Otherwise BTðrX;yÞ ¼ 1.
recommender vehicles in VCN. All false recommendation
trust values obtained from selfish nodes are filtered out
during the indirect trust evaluation. Node ‘X’ enquired
other neighboring vehicles (suppose ‘n’) to send the rep-
utation of node ‘y’. Moreover, the recommendation trust

123
Wireless Networks (2023) 29:2551–2575 2565

4.4 Combined trust (effective node trust) Here,wdirect , wrecomm are the weight coefficients and
evaluation based on analysis, we have used wdirect = wrecomm = 0.5,
means direct trust and indirect trust have equal importance
Combined Trust (or Effective Node Trust) is calculated by in the combined trust calculation. Additionally, the evalu-
Observer Node ‘X’ for the node ‘y’ using the direct and ated combined trust value TðcX;yÞ has been compared with
recommended trust of node ‘y’. The direct trust of node ‘y’ the threshold (i.e., Thc ) for combined trust, and based on the
has been calculated by the Observer node based on own comparison node y is labeled as trustworthy node (T) or
experience with node ‘y’. Moreover, the indirect trust has node y is declared as rogue node (M). The updated status of
been derived from the recommendation from other nodes node y is broadcasted to other vehicles in the network.
about the node ‘y’. Effective trust for node ‘y’ is given by: Finally, the latest combined trust value of node y and trust
TðcX;yÞ ¼ wdirect TðdX;yÞ þ wrecomm TðrX;yÞ ð38Þ status of node y have been updated by all the vehicles in
their routing table. The pseudocode for evaluating the
wdirect þ wrecomm ¼ 1 ð39Þ effective node’s trust and the node’s final status (Rogue or
honest) is presented in Algorithm 2.

123
2566 Wireless Networks (2023) 29:2551–2575

Secure dissemination of critical information e.g., acci- resulting in a secure and dependable communication net-
dent alert, congestion alert, and environmental conditions work between the vehicles. The pseudocode for selecting a
alerts requires set of reliable intermediate relay nodes next-hop relay node to further spread the information is
between sender and receiver. The next relay vehicle is described in Algorithm 3.
selected primarily based on its trust value in this scheme,

123
Wireless Networks (2023) 29:2551–2575 2567

4.5 Temporal complexity analysis T ¼ T ðAlgorithm1Þ þ T ðAlgorithm2Þ þ T ðAlgorithm3Þ

T ¼ OðnÞ þ OðnÞ þ Oðn  nÞ
In the proposed ReTrust model, we have developed and
implemented three algorithms as given above. All the
algorithms are lightweight concerning to the computational
requirements. Algorithm 1 is used to evaluate the recom- 5 Implementation and performance
mendation trust for any vicinal node and to filter out the evaluation
false recommendation values.
This section presents the implementation of the proposed
T ðAlgorithm1Þ ¼ Oðn þ n þ nÞ ¼ OðnÞ
scheme, simulation results and a detailed assessment of the
where, n is the number of recommendations received for obtained results. The ReTrust scheme is implemented using
any particular node. NS-2 and compared with several existing recent security
Algorithm 2 evaluates the effective trust of any node for schemes for VCN such as S-AODV [42], RDA [17], DMN
the observer node. [27], and TBM [32]. Moreover, we have analyzed how the
T ðAlgorithm2Þ ¼ OðnÞ network performs when the number of rogue vehicles
increases and the node’s maximum speed increases in the
where n is the number of neighbours of the observer node communication network scenario. However, the primary
at the particular time instant. objective of the presented simulation was to evaluate the
In the presented algorithm 3, steps 3 through 14 are proposed scheme’s security and efficiency performance in
repeated under the single for loop. This for loop is repeated the presence of rogue nodes. We considered several
n times, where n is the number of neighbouring nodes parameters to exhibit the security and efficiency perfor-
present in the network. Step 5 compares the evaluated mance of the ReTrust scheme and to compare the obtained
decision parameter for all the neighbouring nodes, which results with existing schemes. The performance parameters
requires n comparisons in the worst case. Therefore, the are:
time complexity of the given algorithm 3 is Oðn  nÞ,
• Detection rate is the ratio of correctly identified as
where n is the number of neighbours of the source node.
rogue vehicles over the total vehicles.
The time complexity for algorithm 3 is as follows:
• False positive rate (FPR) denotes the percentage of
T ðAlgorithm3Þ ¼ Oðn  nÞ honest vehicles incorrectly identified as rogue vehicles.
The overall time complexity can be evaluated as: • Successful packet delivery ratio (SPDR) represents the
routing efficiency of the vehicular network. It can be
measured as the ratio of the data packets received by the

123
2568 Wireless Networks (2023) 29:2551–2575

destination node to the data packets sent by the source Table 3 Parameters for simulation in NS-2
node. Simulation parameters Values used in simulation
SPDR ¼ NR =NS ð40Þ
Network Simulator NS-2
Network Area 2000 9 2000 m2
Here, NR and NS represents the number of data packets
Number of vehicles 300
received and sent by receiver node and source node
Number of rogue vehicles 5–50
respectively.
% Of rogue nodes 10% to 50%
• Transmission delay (Td) can be measured in terms of Simulation Time 600 s
the time taken by the data packets to reach from source Routing Protocol AODV, GPSR, ReTrust
to destination. MAC Protocol IEEE 802.11p
E2ED ¼ TD  Ts ð41Þ Packet Size 256 bytes
Packet rate 5 packets/s
Here, TD = Time at destination, TS = Time at source. Maximum Speed 0 m/s to 40 m/s
Simulation Scenario Highway
• Routing overhead/Control overhead: This parameter Maximum Range 250 m
denotes the overhead of using extra routing or control Beacon message size 350 bytes
packets during the transmission process. Number of lanes 04
• Packet loss rate is the ratio of the lost packets to the
T dh 0.4
overall packets disseminated across a communication
T rh 0.2
channel.
• Average Throughput is the average rate at which
messages are successfully disseminated through a
communication medium.
• Detection time is the time required by the observer
5.2 Result evaluation
node to analyse the network traffic and determine the
object vehicle’s trustworthiness.
Using the metrics stated above, the performance of the
• Average hop count is the number of hops required to
ReTrust scheme is compared to that of recent trust-based
disseminate the message from the source to the
security schemes such as S-AODV, DMN, TBM, and
destination node.
RDA. The attacks mentioned above, such as packet drop
A detailed discussion of the simulation environment and attacks and false recommendation attacks, are introduced
result evaluation is given in the following subsections. into the simulation process to evaluate the various security
parameters, and the impact of these attacks is investigated.
5.1 Simulation environment and mobility model Various network characteristics are adjusted to assess the
proposed and existing schemes’ performance, such as the
The Simulation of the proposed ReTrust scheme and number of rogue vehicles (from 5 to 50), the number of
comprehensive performance evaluation is accomplished vehicles (from 20 to 300), and maximum vehicle speeds (5
with network simulator NS-2 [43]. NS-2 is a discrete event to 40 m/s). The selected S-AODV scheme presented a
simulator mostly preferred by eminent researchers in order secure routing mechanism for ad-hoc network centred on
to simulate various realistic vehicular communication AODV routing algorithm. The AODV routing technique is
networks scenarios. The mobility model is used to produce the foundation for a number of existing secure routing
network traffic that is as realistic as possible and delivers solutions. Similarly, the DMN framework offered the direct
significant accuracy in the findings. The considered high- trust-based security model, with verifier nodes for the IoV
way scenario with a straight bidirectional four-lane road network. In addition, the RDA method suggested a trust-
covers a 2000 * 2000 m2 simulation area. Vanet-Mobisim based approach for securing the vehicular network without
is used to generate simulated traffic that closely resembles the need for recommendations or consistency checks. The
real traffic [44]. We have conducted the simulation major reasons for selecting the above-discussed schemes
experimentation 50 rounds, and Simulation analysis is the for comparative assessment are the similar solution
average of the total simulation results. Table 3 summarizes approaches and addressing the identical set of problems in
the many critical parameter settings used in the simulation vehicular networks. Moreover, all the schemes are based
(Table 3). on a trust-based evaluation approach to detect malicious

123
Wireless Networks (2023) 29:2551–2575 2569

activities in the VCN and establish a secure communication DMN increases more rapidly compared to the RDA and
network. ReTrust schemes when the number of rogue vehicles
The performance evaluation subsection is divided into increases in the network. The ReTrust scheme handles the
three parts: (i) impact of rogue vehicles density variations increasing density of rogue vehicles in a better way with
on network performance, (ii) influence of increasing efficient detection and dissemination method. Moreover,
vehicles’ density on network performance, and (iii) impact the ReTrust scheme establishes secure and stable commu-
of vehicles’ speed variations on network performance. The nication paths for information dissemination.
several important results that were obtained are discussed The performance of the security aspect of the proposed
in the below subsections. work in terms of detection rate and the false positive rate is
shown in Fig. 6 compared to other schemes. Fig-
5.2.1 Result analysis with an increasing number of rogue ure 6(a) demonstrates that the proposed scheme has a
nodes better detection rate at a high density of rogue vehicles.
Initially, all the schemes show a similar detection rate at
As shown in Fig. 5(a), the delivery ratio in S-AODV, low rogue vehicle’s density. It can be observed from the
DMN, and TBM schemes decline significantly with the result that the detection rate fell for all schemes when the
increasing rogue vehicles density. Moreover, S-AODV and rogue vehicle ratio in the VCN is increased. The proposed
DMN schemes are based on direct trust evaluations and approach exhibits a 93% detection rate at 50% of the rogue
cannot accurately determine all kinds of selfish behaviours. vehicle density. In addition, the RDA scheme offers a
Similarly, TBM and RDA algorithms are based on direct similar detection accuracy to that of the proposed tech-
and indirect trust evaluation using verifier nodes. The nique. However, the DMN, TBM, and S-AODV are not as
proposed ReTrust scheme exhibits a better packet delivery accurate as of the proposed technique in terms of detection
ratio due to the consideration of only true recommenda- rate.
tions while evaluating the trust of vehicles. Moreover, the Similarly, Fig. 6(b) exhibits the impact of increasing
ReTrust scheme performs superior to other algorithms in rogue vehicle density on the detection method. In terms of
terms of packet delivery fraction due to selecting only the FPR, all strategies perform nearly identically at lower
trusted vehicles in the path from source to destination to rogue vehicle ratios. The false detection rate increases
relay the data packets. significantly as the number of rogue vehicles in the net-
Figure 5(b) demonstrates the impact of increasing rogue work grows. The FPR of the ReTrust scheme is almost half
vehicle density on the packet loss rate. Increasing rogue that of the S-AODV, DMN, and TBM systems. Because of
vehicle density leads to more packet losses and frequent the hybrid trust model, ReTrust and RDA methods can
route disconnections in the network. Moreover, the pres- more effectively identify rogue vehicles in the network.
ence of rogue vehicles increases the number of hops counts However, the Recommendation filtration mechanism pro-
in the communication path used to transmit the data. To vides more accuracy to the ReTrust scheme.
deal with this, detection and routing algorithms must be
able to cope with frequent packet losses and unsta-
ble routes. The packet loss rate for S-AODV, TBM, and

Fig. 5 Comparison of results with respect to the increasing number of malicious nodes

123
2570 Wireless Networks (2023) 29:2551–2575

Fig. 6 Comparison of results with respect to the increasing number of malicious nodes

5.2.2 Result analysis for the density variation vehicle density due to the large number of recommenda-
of the vehicles tions received. However, all the schemes deal with high
detection time due to increased network maintenance at
The impact of varying vehicle density on the network is high vehicle density.
explored using factors including detection time, commu- Figure 7(b) shows the variation in the communication
nication overhead, average delay, and average hop count overhead concerning the increasing number of moving
(Fig. 7). The proposed scheme is compared to other vehicles in the vehicular communication network. The
existing methods based on detection time which is depicted increasing numbers of mobile nodes also increase the
in Fig. 7(a). It is observed that when the vehicle density communication overhead due to the broadcasting of more
increases in the network, the time required for the detection information. Operations related to trust recommendations
of rogue vehicles for all schemes increases. Moreover, dissemination and filtration increases computational and
according to the results displayed the DMN, and TBM communication complexity. The trust computation in RDA
schemes require a significant amount of time to detect the is more complicated than in the other schemes. The rec-
rogue vehicle compared to RDA and ReTrust. The DMN ommendation filtration technique in our proposed method
and TBM schemes take more time at the lower vehicle added some overhead. However, we preferred the binary
density due to the selection of verifier and observer nodes, mode of information transmission to reduce the commu-
respectively. Besides, the RDA and ReTrust scheme takes nication overhead as much as possible. The ReTrust
comparatively more time to detect rogue vehicles at higher scheme demonstrates low communication overhead with

Fig. 7 Comparison of results with respect to the increasing number of moving vehicles nodes

123
Wireless Networks (2023) 29:2551–2575 2571

the better detection rate of rogue vehicles compared to S-AODV scheme is marginally lower at lower vehicle
other methods. However, at high vehicle density to main- density due to the low availability of connections that
tain the high level of security, DMN and TBM attain high increase the hop count in the required path. The ReTrust
communication overhead due to the identification of spe- scheme outperforms other schemes in terms of average hop
cial vehicles among the numerous existing choices. counts once a sufficient number of connected nodes are
The impact of increasing vehicle numbers on average available at a greater vehicle density.
communication delay is depicted in Fig. 8(a). The end-to-
end delay should be kept to a minimum for optimal net- 5.2.3 Result analysis for the mobility variations of nodes
work performance. During the deployment of a security
scheme in the VANET, timely delivery of information Data transmission becomes more difficult as the vehicle’s
packets is a top priority. In the delay evaluation process, mobility grows in the vehicular communication network.
the S-AODV, DMN, RDA, TBM and ReTrust This section compares the simulation results for varied
scheme demonstrates the lower delay at the higher number vehicle mobility for the proposed method with chosen
of mobile nodes in the vehicular network. At high vehicle existing schemes to demonstrate the impact of vehicle
density, more connections are available to disseminate the mobility. The maximum count for the rogue vehicles is
data. Simulation result for delay exhibits that the ReTrust fixed to 20% in the presented scenario. Figure 9(a) shows
scheme incurs lower dissemination delay than the other the result for successful packet delivery ratio with the
schemes. varying maximum speed of the vehicles. The network
Furthermore, Fig. 8(b) shows the influence of vehicle topology changes rapidly when nodes move faster, result-
density on average hop counts required to disseminate the ing in more path breakage and a lower successful packets
network’s data. Average hop counts represent the average delivery rate. The packet delivery ratio for S-AODV
number of hops required in the path to disseminate the decreases very sharply, but the degree of decrement for
message between sender and receiver when both are not in other algorithms is less. The ReTrust scheme shows the
the common transmission range. The hop counts directly lowest decrement during high mobility in the packet
affect the network’s data transfer delay and communication delivery ratio compared to the other existing algorithms.
overhead. Lower hop count in the communication path Figure 9(b) shows that the time delay increases with the
offers faster data delivery with low overhead. Fig- increase in the maximum speed of vehicles. The impact of
ure 8(b) demonstrates that the average hop count decreases mobility variation is more on the S-AODV scheme in terms
when the number of vehicles increases due to the avail- of delay than the other schemes. An increase in speed
ability of more trusted vehicles along with the destination changes the network topology very rapidly, resulting in
vehicle. However, because of the restricted number of frequent path breakage in the network. DMN scheme also
available connections to disseminate the data, average hop shows more delay than the RDA and ReTrust methods.
counts initially tend to rise. ReTrust scheme shows a lower Furthermore, varying mobility has the least impact on the
average hop count in the higher vehicle density compared ReTrsut scheme, and it outperforms existing techniques in
to other existing schemes. Nevertheless, the delay in the terms of delivery performance.

Fig. 8 Comparison of results with respect to the increasing number of mobile nodes

123
2572 Wireless Networks (2023) 29:2551–2575

Fig. 9 Comparison of results as the maximum speed of the nodes increases

Control overhead at lower mobility in DMN and RDA is higher vehicle mobility as these schemes cannot handle the
more significant than in S-AODV and ReTrust, as shown in path instability efficiently.
Fig. 9(c). At low vehicle speed, the control overhead in
S-AODV is less than the other algorithms because, in other
existing models, a trust-based security algorithm increases 6 Conclusion and future scope
the computational complexity. Higher mobility of vehicles
leads to more unstable routes and higher path maintenance. Traditional vehicular ad-hoc networks are unable to detect
Initially, the lightweight ReTrust scheme shows less con- attacks and malicious behavior in the network efficiently.
trol overhead than the RDA algorithms but shows more Furthermore, the presence of rogue nodes can significantly
overhead in contrast to the basic S-AODV and DMN. degrade the performance of the IoV network. Identifying a
Figure 9(d) exhibits the average throughput performance rogue node and establishing trust-based secure communi-
with vehicular mobility variation. It can be inferred from cation is a challenging task in a dynamic vehicular net-
the figure that the higher the speed of vehicles in the net- work. In this work, we have presented a lightweight
work, the lower the network’s throughput performance. security scheme to ensure the security and reliability of the
More unstable routes emerge from increasing mobility, communication network. The proposed algorithm uses the
resulting in increased routing overhead and path mainte- combined trust establishment model to calculate the trust
nance. The ReTrust scheme obtained better throughput value of all the intermediate relay vehicles between source
performance compared to other existing methods for higher and destination. Moreover, this scheme evaluated the trust
vehicle mobility, as the key focus of the ReTrust scheme is of the nodes present in the network based on reliability and
to establish a secure and stable path in the network. The recommendation assessment in the communication net-
S-AODV and TBM scheme show lower throughput for the work. The false recommendations are identified using a

123
Wireless Networks (2023) 29:2551–2575 2573

consistency and similarity measure-based approach. How- hoc networks. In International conference on network and system
ever, this scheme checks the reliability of the node before security. pp. 94–108. Springer, Berlin, Heidelberg.
10. Hasrouny, H., Samhat, A. E., Bassil, C., & Laouiti, A. (2019).
accepting the recommendation from the node. The network Trust model for secure group leader-based communications in
simulator NS-2 was used to complete the suggested sche- VANET. Wireless Networks, 25(8), 4639–4661.
me’s realistic implementation and performance evaluation. 11. Gazdar, T., Belghith, A., & Abutair, H. (2017). An enhanced
After analyzing the results, we can conclude that our distributed trust computing protocol for VANETs. IEEE Access,
6, 380–392.
suggested technique enhances network performance to 12. Fysarakis, K., Hatzivasilis, G., Manifavas, C., & Papaefstathiou,
some level while consuming relatively few network I. (2016). RtVMF: A secure real-time vehicle management
resources. Moreover, the proposed scheme ensures high framework. IEEE Pervasive Computing, 15(1), 22–30.
detection rate, low FPR, high delivery rate, and lower 13. Dua, A., Kumar, N., & Bawa, S. (2014). A systematic review on
routing protocols for vehicular ad hoc networks. Vehicular
communication delay. Communications, 1(1), 33–52.
In future work, the proposed work can be expanded to 14. Kerrache, C. A., Lagraa, N., Calafate, C. T., & Lakas, A. (2017).
improve the precision of our trust model. We plan to TFDD: A trust-based framework for reliable data delivery and
incorporate event authentication along with trust estab- DoS defense in VANETs. Vehicular Communications, 9,
254–267.
lishment in the trust-based framework. Furthermore, we 15. Zhao, H., Sun, D., Yue, H., Zhao, M., & Cheng, S. (2018).
suggest using the fuzzy logic idea to improve the accuracy Dynamic trust model for vehicular cyber-physical systems.
of our presented algorithm’s detection probability. We also International Journal of Network Security, 20(1), 157–167.
recommend that trust establishment with authentication 16. Hu, H., Lu, R., Zhang, Z., & Shao, J. (2016). REPLACE: A
reliable trust-based platoon service recommendation scheme in
schemes be used to protect the network from both insider VANET. IEEE Transactions on Vehicular Technology, 66(2),
and outsider attackers. Moreover, we plan to add some 1786–1797.
additional security metrics to increase the robustness. 17. Yao, X., Zhang, X., Ning, H., & Li, P. (2017). Using trust model
to ensure reliable data acquisition in VANETs. Ad Hoc Networks,
55, 107–118.
18. Soleymani, S. A., Abdullah, A. H., Zareei, M., Anisi, M. H.,
Vargas-Rosales, C., Khan, M. K., & Goudarzi, S. (2017). A
References secure trust model based on fuzzy logic in vehicular ad hoc
networks with fog computing. IEEE Access, 5, 15619–15629.
1. Cui, J., Liew, L. S., Sabaliauskaite, G., & Zhou, F. (2019). A 19. Xia, H., Jia, Z., Li, X., Ju, L., & Sha, E. H. M. (2013). Trust
review on safety failures, security attacks, and available coun- prediction and trust-based source routing in mobile ad hoc net-
termeasures for autonomous vehicles. Ad Hoc Networks, 90, works. Ad Hoc Networks, 11(7), 2096–2114.
101823. 20. Wei, Z., Tang, H., Yu, F. R., Wang, M., & Mason, P. (2014).
2. Kerrache, C. A., Lagraa, N., Hussain, R., Ahmed, S. H., Security enhancements for mobile ad hoc networks with trust
Benslimane, A., Calafate, C. T., Cano, J. C., & Vegni, A. M. management using uncertain reasoning. IEEE Transactions on
(2018). TACASHI: Trust-aware communication architecture for Vehicular Technology, 63(9), 4647–4658.
social internet of vehicles. IEEE Internet of Things Journal, 6(4), 21. Shankaran, R., Varadharajan, V., Orgun, M. A., & Hitchens, M.
5870–5877. (2009). Context-aware trust management for peer-to-peer mobile
3. Hu, X., Chu, T. H., Chan, H. C., & Leung, V. C. (2013). Vita: A ad-hoc networks. In 2009 33rd Annual IEEE International
crowdsensing-oriented mobile cyber-physical system. IEEE Computer Software and Applications Conference. vol. 2,
Transactions on Emerging Topics in Computing, 1(1), 148–165. pp. 188–193. IEEE.
4. Hasrouny, H., Samhat, A. E., Bassil, C., & Laouiti, A. (2017). 22. Chuang, M. C., & Lee, J. F. (2013). TEAM: Trust-extended
VANet security challenges and solutions: A survey. Vehicular authentication mechanism for vehicular ad hoc networks. IEEE
Communications, 7, 7–20. Systems Journal, 8(3), 749–758.
5. Jiang, J., Han, G., Wang, F., Shu, L., & Guizani, M. (2014). An 23. Li, W., & Song, H. (2015). ART: An attack-resistant trust man-
efficient distributed trust model for wireless sensor networks. agement scheme for securing vehicular ad hoc networks. IEEE
IEEE Transactions on Parallel and Distributed Systems, 26(5), Transactions on Intelligent Transportation Systems, 17(4),
1228–1237. 960–969.
6. Tripathi, K. N., Jain, G., Yadav, A. M., & Sharma, S. C. (2021). 24. Fan, N., & Wu, C. Q. (2019). On trust models for communication
Entity-Centric Combined Trust (ECT) Algorithm to Detect security in vehicular ad-hoc networks. Ad Hoc Networks, 90,
Packet Dropping Attack in Vehicular Ad Hoc Networks 101740.
(VANETs). In Next Generation Information Processing System. 25. Kerrache, C. A., Lagraa, N., Calafate, C. T., Cano, J. C., &
pp. 23–33. Springer, Singapore. Manzoni, P. (2016). T-VNets: A novel trust architecture for
7. Huang, Z., Ruj, S., Cavenaghi, M. A., Stojmenovic, M., & Nayak, vehicular networks using the standardized messaging services of
A. (2014). A social network approach to trust management in ETSI ITS. Computer Communications, 93, 68–83.
VANETs. Peer-to-Peer Networking and Applications, 7(3), 26. Sedjelmaci, H., Senouci, S. M., & Abu-Rgheff, M. A. (2014). An
229–242. efficient and lightweight intrusion detection mechanism for ser-
8. Wan, J., Gu, X., Wang, J., & Chen, L. (2019). A trust vice-oriented vehicular networks. IEEE Internet of Things Jour-
scheme based on vehicles reports of events in VANETs. Wireless nal, 1(6), 570–577.
Personal Communications, 105(1), 121–143. 27. Khan, U., Agrawal, S., & Silakari, S. (2015). Detection of
9. Gurung, S., Lin, D., Squicciarini, A., & Bertino, E. (2013). malicious nodes (DMN) in vehicular ad-hoc networks. Procedia
Information-oriented trustworthiness evaluation in vehicular ad- Computer Science, 46, 965–972.

123
2574 Wireless Networks (2023) 29:2551–2575

28. Gharib, M., Lollini, P., & Bondavalli, A. (2017). Towards an Springer Nature or its licensor (e.g. a society or other partner) holds
approach for analyzing trust in cyber-physical-social systems. In exclusive rights to this article under a publishing agreement with the
2017 12th System of Systems Engineering Conference (SoSE). author(s) or other rightsholder(s); author self-archiving of the
pp. 1–6. IEEE. accepted manuscript version of this article is solely governed by the
29. Yan, X., Gu, X., Wang, J., Wan, J., & Chen, L. (2021). A kind of
terms of such publishing agreement and applicable law.
event trust model for VANET based on statistical method.
Dr. Kuldeep Narayan Tripathi is
Wireless Personal Communications, 118(1), 489–503.
Currently working as an assis-
30. Shabut, A. M., Dahal, K. P., Bista, S. K., & Awan, I. U. (2014).
tant professor in Computer Sci-
Recommendation based trust model with an effective defence
ence and Engineering
scheme for MANETs. IEEE Transactions on Mobile Computing,
department at Madhav Institute
14(10), 2101–2115.
of Technology and Science
31. Sharef, B., Alsaqour, R., Alawi, M., Abdelhaq, M., & Sun-
(MITS), Gwalior, India. He has
dararajan, E. (2018). Robust and trust dynamic mobile gateway
completed his Bachelor’s
selection in heterogeneous VANET-UMTS network. Vehicular
degree (B.E.) in computer sci-
Communications, 12, 75–87.
ence and engineering from
32. Tripathi, K. N., & Sharma, S. C. (2020). A trust based model
RGPV Bhopal. Then he
(TBM) to detect rogue nodes in vehicular ad-hoc networks
obtained his Master’s degree
(VANETS). International Journal of Systems Assurance Engi-
(M. Tech) in computer science
neering and Management, 11, 426–440. https://doi.org/10.1007/
and engineering from ABV-
s13198-019-00871-0
IIITM Gwalior, India. He has
33. Baiad, R., Alhussein, O., Otrok, H., & Muhaidat, S. (2016).
received PhD in Computer Science and Engineering from the Indian
Novel cross layer detection schemes to detect blackhole attack
Institute of Technology Roorkee, India. His main research interest
against QoS-OLSR protocol in VANET. Vehicular Communica-
includes Network Security, Machine learning, Internet of Things,
tions, 5, 9–17.
Wireless ad-hoc network, fog computing, and blockchain technology.
34. Kerrache, C. A., Calafate, C. T., Cano, J. C., Lagraa, N., &
He has published various research articles in various peer-reviewed
Manzoni, P. (2016). Trust management for vehicular networks:
journals and conferences.
An adversary-oriented overview. IEEE Access, 4, 9293–9307.
35. Ahmad, F., Kurugollu, F., Kerrache, C. A., Sezer, S., & Liu, L.
Ashish Mohan Yadav has com-
(2021). NOTRINO: A NOvel hybrid TRust management
pleted B.E. degree in computer
scheme for INternet-Of-vehicles. IEEE Transactions on Vehicu-
science and engineering from
lar Technology, 70(9), 9244–9257.
RGPV Bhopal. M. Tech in
36. Liang, W., Long, J., Weng, T. H., Chen, X., Li, K. C., & Zomaya,
computer science and engineer-
A. Y. (2019). TBRS: A trust based recommendation scheme for
ing from RGPV Bhopal, India.
vehicular CPS network. Future Generation Computer Systems,
He has received PhD in Com-
92, 383–398.
puter Science and Engineering
37. Shao, K., Luo, F., Mei, N. X., & Liu, Z. T. (2012). Normal
from the Indian Institute of
distribution based dynamical recommendation trust model.
Technology Roorkee, India.
Ruanjian Xuebao/Journal of Software, 23(12), 3130–3148.
Currently, he is working as an
38. Yang, N. (2013). A similarity based trust and reputation man-
assistant professor in Computer
agement framework for vanets. International Journal of Future
Science and Engineering
Generation Communication and Networking, 6(2), 25–34.
department, School of Comput-
39. Sulaiman, N. H., & Mohamad, D. (2012). A Jaccard-based sim-
ing Science and Engineering at
ilarity measure for soft sets. In 2012 IEEE Symposium on
VIT University Bhopal, India. His main research interest includes
Humanities, Science and Engineering Research. pp. 659–663.
cloud computing, fog computing, and IoT.
IEEE.
40. Asuero, A. G., Sayago, A., & Gonzalez, A. G. (2006). The cor-
Surendra Nagar is currently
relation coefficient: An overview. Critical Reviews in Analytical
working as Assistant Professor
Chemistry, 36(1), 41–59.
in department of computer sci-
41. Mejri, M. N., & Ben-Othman, J. (2016). GDVAN: A new greedy
ence and engineering, Netaji
behavior attack detection algorithm for VANETs. IEEE Trans-
Subhas University of Technol-
actions on Mobile Computing, 16(3), 759–771.
ogy (NSUT), Delhi. He com-
42. Arya, K.V. and Tripathi, K.N. (2013). Power aware and secure
pleted his M. Tech. (CSE) and
routing in mobile and ad-hoc networks. In 2013 IEEE 8th
Ph. D (IT) degrees from ABV-
international conference on industrial and information systems,
Indian Institute of Information
477–482 IEEE.
Technology and Management
43. A discrete event simulator ns-2, (2017), https://www.isi.edu/
(ABV-IIITM), Gwalior, India in
nsnam/ns/.
2013 and 2022, respectively.
44. Husain, A., & Sharma, S. C. (2015). Simulated analysis of
His research interests include
location and distance based routing in VANET with
Image Processing, Computer
IEEE802.11p. Procedia Computer Science, 57, 323–331.
Vision, Image Super-resolution,
Face Hallucination, WSNs, Ad-hoc Networks, Machine Learning, and
Publisher’s Note Springer Nature remains neutral with regard to Recommendation Systems.
jurisdictional claims in published maps and institutional affiliations.

123
Wireless Networks (2023) 29:2551–2575 2575

S. C. Sharma has completed M.

Sc and M. Tech degree in elec-
tronics and communication
engineering and PhD in elec-
tronics and computer engineer-
ing from IIT Roorkee, India.
Currently, he is working as a
professor in IIT Roorkee. He
has more than 25 years research
and teaching experience. He has
published over two hundred
research papers in national and
international journals and con-
ferences. His research interest
includes cloud computing,
wireless sensor networks, and IoT.

123