A Trust Management Framework for Vehicular Ad Hoc Networks

International Journal of Security, Privacy and Trust Management (IJSPTM) Vol 12, No 1, February 2023
DOI: 10.5121/ijsptm.2023.12102 15
A TRUST MANAGEMENT FRAMEWORK FOR
VEHICULAR AD HOC NETWORKS
Rezvi Shahariar and Chris Phillips
School of Electronic Engineering and Computer Science, Queen Mary, University of
London, London, UK
ABSTRACT
Vehicular Ad Hoc Networks (VANETs) enable road users and public infrastructure to share information
that improves the operation of roads and driver experience. However, these are vulnerable to poorly
behaved authorized users. Trust management is used to address attacks from authorized users in
accordance with their trust score. By removing the dissemination of trust metrics in the validation process,
communication overhead and response time are lowered. In this paper, we propose a new Tamper-Proof
Device (TPD) based trust management framework for controlling trust at the sender side vehicle that
regulates driver behaviour. Moreover, the dissemination of feedback is only required when there is
conflicting information in the VANET. If a conflict arises, the Road-Side Unit (RSU) decides, using the
weighted voting system, whether the originator is to be believed, or not. The framework is evaluated
against a centralized reputation approach and the results demonstrate that it outperforms the latter.
KEYWORDS
VANET, Trust Management, Security, Tamper Proof Device, Malicious Behaviour
1. INTRODUCTION
Vehicular Ad Hoc Networks (VANETs) can provide traffic advice, safety announcements, and
infotainment services to road users. Typically, a vehicle may report an emergency event with
other road users or may request the location of a petrol pump or a nearby parking area. VANETs
are also deployed to mitigate the aftereffect of road incidents and to warn vehicles in advance.
However, as this application involves wireless communication, it is at risk of security attacks.
Additionally, drivers can fraudulently broadcast false messages. To be successful, messages must
be accurate and trustworthy, otherwise, with a malicious untrue message, a vehicle can mislead
many others causing congestion or other undesirable phenomena.
Figure 1. Typical Mechanism for Reporting an Accident

16
For example, as shown in Figure 1, suppose a vehicle “V” broadcasts a message reporting a
crash. This message could be truthful or false. If other vehicles receive a false message, their
subsequent detour will impact their travel time. However, for a truthful announcement, the detour
permits them to avoid potential congestion.
In a VANET, outsider attacks can be thwarted using a cryptographic scheme but not insider
attacks. Trust-based approaches are used to thwart insider attacks from malicious authorized
users [1, 2, 3]. It is noted in [4, 5] that trust schemes can improve security by identifying
dishonest vehicles and revoking messages from them. Even so, trust approaches cannot protect
VANETs completely [1]. Basically, the trust that vehicle W attributes to vehicle V is the
confidence W places in a set of actions from V. Typically, the reliability of relayed information is
periodically evaluated using predefined metrics and computational methods [6]. Vehicles that
consistently maintain a good trust score can be considered trustworthy by others as their current
trust scores rely on their previous trustworthy announcements. However, it is not guaranteed that
a trusted vehicle will always broadcast trustworthy messages.
In existing approaches [2, 3, 6-8] both trusted and untrusted vehicles can broadcast messages.
Untrusted vehicles are expected to broadcast more malicious messages than trusted vehicles,
which produces an additional demand on the network both in terms of message volume and the
verification process. This places a considerable burden on the receivers. Methodologies based on
direct and indirect trust require regular monitoring of activities across both single and multi-hop
transmission ranges. Some approaches [2, 4, 9] can result in considerable trust metric exchanges
to verify the original announcement. These messages, along with the event announcement,
complicate the situation as it is necessary to evaluate their validity in a short time frame due to
fast vehicle movement [10]. The authors in [6] claim receivers should decide the trust of
messages in a short timeframe. However, when receivers independently compute trust from their
neighbours’ trust metrics, they suffer from high response times [2, 6, 9]. Alternatively,
approaches that allow trust computation at a centralized server need to communicate to obtain
updated trust information concerning the sender vehicle. This introduces an additional delay in
the decision-making process concerning emergency events. Consequently, some vehicles may
drive into the event zone despite being previously warned, as suggested in [11]. Also, there is an
open debate [4, 8] regarding how often a centralized server should communicate revised trust
data. Therefore, this paper proposes a novel Tamper-Proof Device (TPD)-based sender-side trust
management framework for VANETs with the following features:
 To the best of our knowledge, for the first time, this framework employs a sender-side trust
model to regulate access control using information accuracy, delay, and positional
differences collected from the sender vehicle itself. Unlike other approaches [2, 4, 7, 8, 10,
12, 13], there is no flow of trust metrics unless a reporter vehicle refutes an announcement.
 Various classes of messages along with their associated trust threshold are defined for
regulating access control which confirms that only trusted vehicles can announce messages.
 The scheme employs a collaborative untrue message discovery algorithm for detecting
various forms of attack.
 Receivers can instantly act on a sender’s message knowing that the sender must have
sufficient trust, or they can report an untrue attack if they do not see the event on the roads.
 The framework is simulated in the Veins to verify its satisfactory operation. The
communication overhead and response time are compared against a typical reputation
approach [8] with varying vehicular densities and speeds. Moreover, the accuracy of the
framework is presented with differing percentages of malicious and benevolent feedback.

17
Our paper is organized as follows. Section 2 reviews existing trust models. Section 3 presents the
core parts of the proposed trust model. Section 4 illustrates the system model and verification of
the framework. Section 5 then analyses the framework and provides a performance comparison
against a baseline approach. Finally, in Section 6, we conclude this work.
2. RELATED WORK
Till now, many trust models for VANETs have been proposed for evaluating message
trustworthiness [5]. These models are categorized into three groups. The first group is called the
Entity-Oriented Trust Model (EOTM) which verifies only an entity’s trust. The second group is
called the Data-Oriented Trust Models (DOTM) which focus on evaluating the trustworthiness of
data only. The third group is called the Hybrid Trust Model (HTM) which evaluates both entity’s
trust and the reliability of data. Below trust models from these categories are highlighted briefly.
Entity-oriented trust models are typified by [7], where the researchers use a Tamper Proof
Module (TPM) on every vehicle to find the cost for the transmission and then adjust trust from
the receivers' feedback. However, the trust score updating leads to excessive communication.
Conversely, the authors in [14] apply fuzzy logic to calculate trust using experience, plausibility,
and location accuracy. It can detect bogus messages and alteration attacks. However, additional
communication is required as vehicles consult with fog nodes to check the location accuracy. Ref
[9] uses both fuzzy logic and Q-learning for trust calculation. This approach is evaluated in terms
of precision and recall with varying numbers of malicious vehicles. However, the model requires
repeated sensing of messages from neighbours. In contrast, Ref [2] uses the Bayesian rule and
Dempster-Shafer Theorem (DST) for trust calculation. It combines independent beliefs to
determine the trust of a vehicle. However, an erroneous recommendation can bias the trust
calculation. Reference [4] uses reputation and receivers’ feedback on received messages to
calculate trust. The scheme is evaluated in presence of false messages in both urban and highway
scenarios. Nevertheless, it may suffer from excessive trust metric dissemination. The researchers
in [15] propose a past interactions-based reputation management scheme for VANETs. A vehicle
collects a signed reputation from the server to attach to its messages. A receiver verifies the
message to accept or reject it and updates the reputation of vehicles on the server. However, this
approach requires periodic reputation exchanges with the server. Another trust model is proposed
in [16]. They consider familiarity, packet delivery ratio, timeliness, and interaction frequency as
parameters to manipulate final trust. This model is analyzed considering the recent history of
interactions but not considering any attacker model. In [17], the authors introduce a blockchain-
based trust management protocol for VANETs to update of trust of vehicles as low, medium, and
high at the Trust Authority (TA) using tamper-proof logs periodically. However, this is not
validated using any known attack.
Data-oriented trust models are typified by [18], where packets are forwarded along the most
trusted path using a trusted routing protocol. An intrusion detection module thwarts only denial of
service (DoS) attacks. Ref [19] presents an infrastructure-less, data-oriented trust model which
verifies content similarity and conflict as well as route similarity. Conversely, the reference [20]
proposes a distance and geolocation-based probabilistic approach to estimate the trust from
received data. This model does not forward a message beyond a certain distance. Ref [21]
proposes a Bayesian Inference-based voting mechanism and vehicles run Dijkstra’s algorithm as
a route update requires upon every message arrival. Messages carry a time parameter as road ID
while being forwarded. However, this requires frequent maintenance of routes. The researchers in
[22] propose a trust model called FACT for achieving reliable information dissemination in
VANETs using two modules. The first module evaluates the trust of the message, whereas the

18
second one finds a highly trusted path for forwarding messages. However, this is an application-
oriented scheme that does not use any infrastructure to monitor activities.
The researchers in [5] evaluated one data-oriented, one entity-oriented, and one hybrid trust
model under various adversary scenarios. In addition, a risk assessment model is also presented
for the identification of critical vulnerabilities. Ref [6] checks the reliability of messages using
direct interactions and stores previous interactions and the trust of neighbours in a local database.
However, there is no false trust message detection scheme. Conversely, the authors in [12]
calculate data trust from multiple vehicle responses and vehicle trust from functional and
recommendation trust. Though this model considers simple, bad mouth and zigzag attacks, but
not compared with any model. In [3], RSUs use hash message authentication code (HMAC) and
digital signature to evaluate the trust of vehicles based on neighbour trust values and only reward.
They measure communication overhead and suggest integrating the ID-based and batch
signatures in the future. The researchers in [8] update vehicle reputations at the RSUs using the
noticed events from vehicles. The RSU next announces updated reputations to vehicles.
Receivers store all the messages about an event until a timer expires to decide whether an event is
true or false. However, this model suffers from high response times and communication
overhead. Alternatively, in [23], a self-organizing trust model is considered for both urban and
rural settings which can detect fake locations, and fake times and revoke messages as necessary.
This model validates the trust of messages and then accepts the message with the highest trust for
an event. However, its efficacy is not analyzed. In [24], the authors present a trust model, where
the entity-centric trust model of this scheme thwart the black-hole attack and selective forwarding
attack. The data-centric trust model is used to discover relations among data and performs trust
evaluation based on utility theory. The data trust model can be further improved by selecting the
appropriate utility parameter. Alternatively, in [25], a risk-based hybrid trust model is proposed
and compared with a multi-facet-based trust model. The result suggests that it always selects a
low-risk action which is different from what the trust-based approach suggests. However, this
work is only suitable for a clustered architecture which is unrealistic for VANET. Ref [26] uses a
Bayesian inference-based direct trust and recommendation trust to calculate the final trust. This
model finds the confidence of direct trust to avoid the costly recommendation trust calculation.
Though this approach is compared with two models, they consider only packet drop and
interception as malicious behaviours.
Many blockchain-based trust models are also present in the literature. Here, we only highlight
two approaches though our proposed model is not blockchain-based. Ref. [27] presents a three-
layer blockchain-based trust model for VANETs using Dirichlet distribution, regression, and
revocation. They consider simple, slander, and strategic attacks along with both normal, and
malicious servers. However, the work does not reward benevolent activities from vehicles. Also,
in [28], the authors present a decentralized blockchain-based trust model which selects a message
evaluator through RSU collaboration. The approach finds the rating for messages, the sender, and
the evaluator. Next, they calculate the global trust of a node based on the rating, and message
quality. They preserve trust data in the blockchain and use a consensus process to insert blocks.
They claim that their approach can prevent Sybil, message spoofing, bad-mouthing, and ballot-
stuffing attacks. However, this model is not compared against other trust-based models.
3. PROPOSED SYSTEM
Ideally, a trust model promoting security should incur little or no extra burden in terms of
computational and communication cost. In a VANET, vehicles typically meet each other
randomly and fleetingly. Thus, there is little time for decision-making based on trust. With a
receiver-side trust model, vehicles with a poor trust score can still send messages although these

19
will typically be ignored by receivers once their trust level is verified. However, this takes time,
so vehicles may unnecessarily experience events such as traffic jams. To this end, our research
proposes a novel sender-side trust management framework that reduces the amount of trust
information passed over a VANET and blocks untrusted transmission attempts. It uses a sender-
side TPD on vehicles to prevent unauthorized access and regulate transmissions based on the
level of trust. Different classes of messages are created that associate different trust thresholds to
permit their broadcast. Drivers improve their trust score from valid announcements, forwarding,
beaconing, and clarifying events to an RSU. Also, a driver builds trust from RSU rewards
whenever “wins a dispute” over another vehicle. The beaconing reward is only given when a
driver is not blocked, and trust is less than 0.5. The TPD also punishes drivers when
announcements are delayed, or the vehicle is travelled more than a threshold distance besides the
arrangement of RSU punishments whenever a driver “loses a dispute” to another vehicle. Since
announcements are regulated by the sender’s TPD, receivers can believe messages and the
sender’s trust instantly.
3.1. System Assumptions
The framework assumes that the security of the TPD is beyond the scope of the proposed
framework as it relates to physical layer protection. Also, we do not consider other security
aspects with this trust framework as we believe that existing security techniques can address
authentication, privacy, and integrity. A security approach that supports these functionalities
could be incorporated with the framework to confirm the authenticity of the driver and/or
messages with other entities, secure the privacy of the driver, and any Hash Message
Authentication Code (HMAC) for achieving integrity. For example, when a driver registers with
the TA, the driver can obtain private and public keys to encrypt and decrypt messages and can
obtain a pseudo-identity associated with his driver ID for securing privacy with other drivers, and
HMAC can be used to maintain integrity [1]. RSUs, official vehicles, and the Trust Authority
(TA) are also considered fully trustworthy. Both the TA and TPDs are governed and owned by
the Road Transport Authority (RTA). The resilience of the TA infrastructure is beyond the focus
of this work. We assume a driver has a built-in dashboard with designated touch buttons to
display the classes of messages available given his/her current trust score and to generate specific
emergency events for other vehicles. Furthermore, a TPD can access GPS to determine the
location of the vehicle.
3.2. Registration, Blocklisting, and Redemption
Drivers may register themselves with the TA directly using an online form with the vehicle plate
number as vehicle ID and driver's license number as driver ID. Since this is an external means,
this is outside the scope of the framework. Alternatively, if the system chooses to send a
registration message from the driver when they start initially; then the RSU forwards the
registration and confirmation of registration to and from the TA.
RSUs send the decisions of disputed events to the TA to store in a driver profile database which
keeps driver and vehicle information, the event information, and the reward/punishment. When
the TA receives a decision on a disputed event, then it searches the driver profile database. If
three malicious events have been found in a limited timeframe, the TA sends a blocking
confirmation message to the RSUs in the driver’s vicinity. When the vehicle receives a blocking
confirmation message, the TPD blocks further access of the driver and acknowledges the
blocking confirmation message with the TA via an RSU. The blocked driver can only
send/receive beacons into the VANET. Additionally, a blocking message can be generated from
the vehicle’s TPD when the driver’s trust score crosses the lowest acceptable trust limit. This
message is forwarded by an RSU to the TA and the same mechanism is followed to block the

20
access of traffic events for the driver in the VANET. By default, based on experimentation,
regular drivers obtain access to traffic events in the VANET with a trust score between 0.06 to
0.9. Whenever the trust score becomes lower than 0.06, an external mechanism requires the
driver to communicate with the TA to obtain redemption from blocking. We assume this is within
the jurisdiction of the road transport authority (RTA) and may involve issuing a monetary penalty
or other sanctions.
3.3. Framework Components
Both regular vehicles and official vehicles are present within the framework. The framework is
extensible so that more vehicle classes could be added. The approach supports the same vehicle
accommodating multiple drivers via individual driver trust management. The actions and
responsibilities of each vehicle type are limited to their role. Every vehicle is pre-equipped with a
built-in On-Board Unit (OBU), comprising a Global Positioning System (GPS) for location
access, a transceiver to communicate with other entities, and a TPD that manages the trust and
regulates transmissions. We define the following actors based on their roles:
 Senders: are drivers that can originate both true and untrue announcements relating to an
incident, such as an accident, subject to their trust score. For a true announcement, the trust
manager within the TPD rewards the driver if the claim is not disputed within a given time.
 Reporters: are drivers that refute an announcement of a sender and receive a reward or
punishment if the challenge is confirmed or dismissed, respectively. If they do decide to make
a report, they may do so either truthfully or falsely. Failure to make a report carries no
penalty.
 Receivers: are drivers that receive messages from any entity and relay them automatically
provided the hop limit is not reached and their trust is sufficient.
 Clarifiers: If a dispute is detected at an RSU, the RSU transmits a query message seeking
clarification concerning a disputed incident. Vehicles that receive this message can choose to
answer the query, i.e., to respond to the RSU, confirming or denying that the incident has
taken place, or ignore it. If they respond, they are considered clarifiers.
 Road-Side Units: are automated units that receive information from senders, reporters, and
clarifiers, either directly or via intermediate vehicles that rebroadcast the received messages. If
information from multiple senders, reporters, or a combination of these conflicts, then the
RSU will rule on the dispute. RSUs act as an intermediary between the vehicles and the TA.
 Trust Authority: is the ultimate authority in this framework which validates registration and
blocklisting of drivers. The TA blocklists a driver whenever it receives a blocklist message
initiated from a TPD or if it finds the three malicious events (3ME) for the same driver within
a configurable timeframe. The TA then replies with a blocking confirmation to the RSUs in
the vicinity of the last disputed event to reach the vehicle’s TPD. Incidents reported by RSUs
are saved by the TA in an incident database including the location, timestamp, and incident
information. The TA also maintains a driver profile database containing the
reward/punishment history of drivers.
 Official Vehicles: This framework considers police, ambulance, and fire service vehicles as
official vehicles. Their primary task is to respond to emergency issues on roads by
cooperating with RSUs. They are always trusted.
3.4. Trust Evaluation Mechanism
This framework sets an initial trust score of 0.45 for regular vehicles to avoid the cold start
problem and with the expectation that they will achieve a trust score of 0.5 relatively quickly so
that they can then announce all events that the framework supports. Regular vehicles are

21
considered trusted when their trust score is 0.5 or higher. A regular vehicle’s trust score cannot
go above 0.9 or below 0.05. The trust T of a regular vehicle i is expressed by Eqn (1).
Ti= {t | t ∈ R | 0.05 < t ≤ 0.9} (1)
The following rules govern the actions of regular vehicles considering T as trust. If T ≤ 0.05
(blocked state) of a driver and/or vehicle, then TPD sends a blocking message automatically to
the TA, and the vehicle can only generate periodic beacons. Then the TPD waits and blocks
network access for all traffic events upon blocking confirmation from the TA. If 0.05 < T ≤ 0.25
(not trusted state), then the vehicle can announce periodic beacons as well but cannot forward
events from others. If 0.25 < T < 0.5 (lowly trusted state), the vehicle can send beacons, make
limited announcements, and can forward events from others. If 0.5 ≤ T ≤ 0.9 (trusted state and
highly trusted when (T = 0.9)), the vehicle can forward and announce all classes of events. If
regular vehicles spread untrue messages multiple times, then they receive incremental
punishments from RSU. If this count becomes three for severe situations like false accident
announcements, then the network access for the driver is blocked. Since the framework considers
official vehicles, they are assigned a higher trust score over regular vehicles (i.e. T = 1.0) as
regular vehicles should not be trusted more than an official vehicles.
We envision a driver’s dashboard as consisting of a set of buttons for supported actions in the
framework. Appropriate buttons can be pressed relevant to a specific type of road incident. There
are three main classes of messages in the hierarchy. The lowest class consists of beacons and
“wave” service announcements, though a blocked driver cannot use the “wave” service facility.
The next class of messages consists of announcements of poor road conditions, debris, road
defects, and so forth. These can only be broadcasted by drivers with a trust score greater than
0.25. The highest class of messages consists of announcements for accidents, traffic jams, road
closures, etc., as well as untrue attack reporting messages. To announce a message from this
class, a driver needs to have a trust score of at least 0.5. Algorithm 1 shows the announcement,
retransmission, relaying, feedback, and reporting activities of regular vehicles. In Algorithm 2,
the TPD trust update and blocking management are shown for regular vehicles. Notations and
symbols for Algorithms 1 and 2 are provided in Table 1.

22
Table 1. List of Notations
Algorithm 1. regular vehicle traffic event management
Input: Driver ID, Vehicle ID, events, trust of drivers, hop and retransmit limit
Output: controlled broadcasting, relaying, reporting, and sending feedback
1. case eventType of
2. witnessed-event:// to warn others.
3. if (Ts (Ds (Vs))≥ ATTL(evte))
4. Ds(Vs) prepares and broadcasts the evte
5. Send metrics to TPD to find
Rewtpd/Puntpd
6. end if
7. reported-event:// to report the received
event.
8. if (Vs decides evte=false) and (Vs visits
event place within Tint ) and (Ts (Ds (Vs))
≥ 0.5)
9. Send untrue(evte) towards RSU
10. Notify TPD to add Ts= Ts + Rewardunt-
atck
11. end if
12. relayed-event:// to relay event up to hop
limit.
13. if (Vs gets an evte or an untrue(evte)
from a Vrep first time)
14. if (Vs sends evte or untrue (evte))
15. Return
16. end if
17. if (Ts ∈ (Ts > 0.05 and Ts <=0.25)
18. Send a LowTrustmsg
19. else
20. if TTL(evte or untrue(evte) ≥ HL)
21. Stop resending evte or untrue(evte)
22. else
23. Resend evte or untrue(evte) up to
HL
24. Notify TPD to add Ts= Ts
+Rewardf
25. end if
26. end if
27. end if
28. retransmit-event: // to repeat the
broadcasting
29. if (no_of_time ≤ RTL)
30. Resend evte
31. end if
32. feedback-event: // to send feedback.
33. if (Vs receives a RSUquery about evte)
34. if (Vs is a Vrep or is the sender of evte)
35. Return
36. end if
37. if TTL(RSUclarif_query) < HL
38. Resend RSUclarif_query message
39. end if
40. if (Vs visits the event location within
Tdis)
41. Send feedback
Notation Meaning Notation Meaning
RSUr rth
RSU HL and RTL hop limit and retransmission
limit
Ts(Ds(Vs) trusts of driver s of vehicles Rewardf,
Rewardclar, and
Rewardunt-atck
reward for forwarding,
clarification, and reporting
Vrep, Vrec, & Vtrust-
cla
reporter, receiver, and trusted
clarifier vehicle
LowTrustmsg forwarding is not possible
with low trust
timerreward-withhold when to process
reward/punishment
timerbilst to check blocking condition
evte and
untrue(evte)
traffic event and reporting the
evte
driver_List registered driver list
RSUclarif_query clarification query from RSU Trusts, Trustd saved and initial trust
Tdis, andTint time threshold to send
feedback and to report evte
Complaint_List list of reported
announcements
TTL, and Mcls Time-To-Live, class of
messages
longDelayed driver delayed than the upper
limit
ATT(Mcls) associated trust threshold of
Mcls
Msgblock &
Msgblock-conf
blocking and blocking
confirmation message
Rewr/Punr and
Rewtpd/Puntpd
reward/punishment from a
RSUr, andTPD
PosDiff distance between the event
and announcement location

23
42. Notify TPD to add Ts = Ts+
rewardclar
43. end if
44. end if
45. end case
Algorithm 2: trust and blocking management at the TPD
Input: Announced evte, reporting status of evte, Msgblock-conf , PosDiff, delay, Rewr/Punr,
Output: Trust update and access blocking.
2. periodic-blocking-checker://
3. if Ds(Vs) is unblocked) and (timerblist
expires) and (Ts(Ds(Vs)) <=0.05))
4. TPDs issue a Msgblock to reach TA.
5. end if
6. if Msgblock-conf comes from TA for Ds(Vs)
7. Disable the network access for Ds
8. end if
9. RSU reward/punishment: // add with
trust.
10. if (Rewr/Punr from an RSUr for the
D(s(Vs)))
11. (Ts(Vs(Ds))= (Ts(Vs(Ds))+Rewr/Punr
12. end if
13. metrics: // Applies reward withholding.
14. if (evte=false by receiving a complaint)
15. rewardevt-e =0
16. else
17. case [PosDiff | D]
18. 0<PosDiff<300m | 0<D<15s:
19. rewardtpd =0.08.
20. 301<PosDiff<500m | 16<D≤30s:
22. 501<PosDiff <800m | 31<D≤60s:
24. 801<PosDiff<1200m | 61<D≤120s:
26. 1201<PosDiff<1500m | 121<D
≤150s:
27. rewardtpd =-0.01.
28. longDelayed=true
29. PosDiff >1500m | D>150s:
30. rewardtpd =-0.05.
31. longDelayed=true
32. end case
33. if longDelayed=true
34. call reward/punishment process
immediately
35. else
36. start timerreward-withhold to process
reward
37. end if
38. end if
39. TPD reward/punishment://add with
trust
40. if (broadcasted msg_id ∈
Complaint_List)
41. Update Ts (Ds (Vs) = Ts (Ds (Vs)
42. Return
43. else
44. Update Ts (Ds (Vs)= Ts (Ds
(Vs)+rewardtpd
45. if (Ts (Ds (Vs))>0.9
46. Update Ts (Ds (Vs)=0.9
47. end if
48. if (Ts (Ds (Vs))≤ 0.05)
49. Update Ts (Ds (Vs)=0.05
50. Start timerblist
51. end if
52. end if
53. complaint-on-broadcasted-event:// save
report.
54. if the broadcasted msg_id has a
complaint
55. Save the complaint into the
Complaint_List
56. end if
57. driver-change-event:// to change driver
58. Extract the driver_name Ds
59. if (Ds exists in the driver_List)60. Use
Ds as the current driver and Trusts
61. else
62. Add Ds in driver_List and Ts=Trustd
63. end if
64. end cas
A potential reward is initially assessed at a TPD and then withheld for a period before adding it to
the current trust. During this time, a TPD checks whether any complaint has been raised by any
reporter. Rewards are calculated based on message accuracy (no complaint), location difference,

24
and delay/responsiveness. Thus, the framework promotes emergency event announcements at the
earliest possible opportunity. The distance a vehicle moves between the event location and the
vehicle’s current position is passed to the TPD to determine the location difference. Delay is
calculated as the difference between the announcement and the observation time on the road. The
TPD uses this information to assess the reward/punishment for the announcement. Also, the
framework suggests a vehicle should not travel more than 500 meters or the next traffic signal to
earn a higher reward from the announcement. The trust Ti is updated inside the TPD using Eqn
(2).
Ti = Ti-1 + Ri “or” Pi (2)
Here, Ti is the revised trust score of a vehicle after adding a reward/punishment to its current
trust Ti-1. Ri “or” Pi is the estimated reward or punishment for the ith
message announcement.
The set of rules used by the TPD for deciding the appropriate reward/punishment magnitude for a
given announcement is shown in the “metrics” event handling in Algorithm 2. In the
reward/punishment assessment, the accuracy of the announcement, delay, and location difference
is considered and then the current trust is updated using the assessed reward/punishment. The
assessed reward is then withheld for a period and the reporting status of the announced message
is checked before the trusted update. If the message is accurate but the driver exceeds the
thresholds, then no reward or a nominal punishment is issued. But the assessed punishment is
deducted immediately from the current trust. However, if the message is inaccurate (i.e., a
complaint received during the reward withholds period), then the driver receives no reward for
the announcement from the TPD and defers the reward/punishment decision to the RSU. As a
VANET is a time-critical system, vehicles should disseminate information promptly. Thus, the
delay and distance traveled are given prime importance in the reward calculation besides the
message accuracy.
3.5. Functional Diagram of the Proposed Framework
In Figure 2, assume a driver sees an incident and wishes to announce it. The framework first
checks the trust score from the TPD and determines if the action is eligible with the driver’s
current trust. If this test is passed, then the driver announces the incident. Other “receivers”
forward it up to the configurable hop limit. As this is an original announcement, the driver is
classified as a "sender". If this announcement reaches subsequent drivers who visit the same
location later, they can notice whether the said event occurs or not. However, if any driver
believes the announcement to be untrue, that driver can send a complaint to the RSU. When the
RSU receives this complaint, the RSU requests “trusted clarifiers” to respond, confirming or
denying the claim.
It collects feedback from these trusted clarifiers who have recently visited the event location.
After this, the RSU rules on the validity of the event and penalizes or rewards the respective
vehicles. An RSU always informs the TA of the outcome of a dispute, which could be a driver
being malicious. It is then up to the TA to check prior behaviour for three malicious activities
from a specific driver over a configurable time and block this driver. The TA sends a blocking
confirmation message to RSUs in the vicinity from where the TA receives the last dispute
decision. These RSUs forward it to the concerned vehicle which receives and acknowledges the
instruction. Alternatively, if a driver’s trust reduces to 0.05, the TPD generates a blocking request
message which a nearby RSU forwards to the TA. Then the TA blocks this driver and informs the
respective TPD via the RSUs in the vicinity of the vehicle.

25
Figure 2. Functional Diagram of Proposed Framework
3.6. RSU Traffic Event Management/Functionality
RSUs always listen to traffic events and share them as necessary based on their severity. RSUs
receive regular beacons from vehicles. In response, RSUs send beacons periodically to notify of
their roadside existence so that other entities can request services. When RSUs receive an
emergency traffic event message from a sender vehicle, they rebroadcast the same towards
neighboring vehicles so that oncoming vehicles whose route includes the problematic road may
avoid it. RSUs also share certain events with nearby RSUs so that vehicles in a greater region
may avoid the problematic road, if appropriate. For some events, RSU will continue to
periodically announce it until they receive notification from an official vehicle to confirm the
event is resolved. When a traffic event occurs and if the RSU receives an AttendingBY-Voff
message from an official vehicle, the RSU can confirm the event has occurred. The RSU
continues to announce the traffic event periodically until the reception of a sorted traffic event
from the official vehicle. When the event sorted message arrives, the RSU stops rebroadcasting
the original traffic event towards oncoming vehicles. Rather it starts broadcasting only the sorted
traffic event up to a retransmission limit as well as forwarding this message to nearby RSUs
based on the severity of the original traffic event. RSUs rebroadcast and forward traffic incidents
to the TA from sender vehicles besides storing traffic incident information until it is resolved.
Each local service point, for example, petrol pumps, and parking is registered in advance with the
nearby RSU. Whenever any vehicle sends any query seeking information regarding any service,
the local RSU sends a reply to the service query containing the information of queried service or
it says it has no information if it does not know.
An RSU assigns a fixed amount of reward and sets the punishment for disputed announcements
using an Incremental Punishment Policy (IPP). An RSU forwards the decision of a disputed event
to the TA. Then TA checks the malicious event count for relevant drivers. If the 3ME condition
holds then the TA sends a blocking confirmation message to the RSUs in the vicinity of the last
disputed event. After that, these RSUs broadcast this message to the vehicle. Besides these
functionalities, an RSU also resolves disputes when untrue attack messages arise which is
illustrated in Section 3.7.
3.7. RSU Untrue Message Detection
If an RSU receives conflicting information from a sender and a reporter, it initiates a
“collaboration” process to determine the validity of the disputed event. To this end, first, an RSU
broadcasts a send-a-reply message to all trusted clarifiers in the vicinity including possible

26
official vehicles and waits for a timer to expire when the feedback collection is finished, as
depicted in Algorithm 3. Notations and symbols for this algorithm are listed in Table 2.
Table 2. List of Notations.
Notation Meaning
evte-sorted sorted event evte
untrueHandledList list of disputed cases which RSUr has decision
send-a-reply(evte) ask Vclas to send feedback on evte
Locreq(servicei) and Locrep(servicei) location of ith
service query and ith
service reply
RSUr{known service} known services at RSUr
timersc timer to collect feedbacks by an RSU
replyoff reply from Voff
untrue_id untrue_attack message id
AttendingBY-Voff (evte) attending evte by a Voff
rewV, punV rewarded and punished vehicle
Algorithm 3: RSU untrue attack handler
Input: untrue attack, feedback message, trust, lists to save events
Output: initiate feedback collection for a timer, find rewarded/punished vehicle
1. while running
3. untrue attack: // deals with untrue
attacks.
4. if unique untrue(evte) from RSUs /Vs
5. Insert into untrueAddedList
6. if untrue(evte) ∈ untrueHandledList
7. Return
8. else
9. Insert into untrueHandledList
10. if untrue(evte) from a Vs
11. Broadcast a send-a-reply(evte)
12. Start a timersc to collect
feedbacks
13. end if
14. end if
15. else
16. RSUr receives an untrue(evte) from
a Voff
17. Call rew-pun-generator(Voff, Vs)
18. end if
19. feedback: // collect all the feedbacks.
20. while (timersc is not expired)
21. if unique feedback fu from Vcla is for
RSUr
22. Insert in vector <f0, f1, … fn>
23. if fu is from a Voff
24. if fu is the same as the Vs’s event
25. Call rew-pun-gen(Vs,Vrep)
26. else
27. Call rew-pun-gen (Vs,Vrep)
28. end if
29. Update rewardList and
punishmentList
30. forwardMsgtoRSUs(decision_untrue)
31. if count(3ME(Vs or Vrep)) ≥ 3
32. Send a Msgblock(count(3ME(Vs
or Vrep) ≥ 3)) to TA
33. end if
34. end if
35. else
36. The feedback is for different RSUs
37. end if
38. end while
39. decision-of-untrue: // to resolve
dispute
40. if timersc expires
41. if the untrue(evte) has a decision
42. Return
43. else
44. Sum=0
45. case feedbackType of
46. Positive: Fi =1
47. Negative: Fi =-1
48. Unsure: Fi =0
49. end case
50. for each Fi from feedback vector
<Fn, Tn>
51. Sum +=Ti*Fi
52. end for
53. if Sum>0
54. Vs send true event, Vrep send
false report
55. Call rew-pun-generator(Vs, Vrep)

27
56. else if (sum<0)
57. Vs send false event, Vrep send
true report
58. Call rew-pun-gen (Vrep, Vs)
59. else
60. Undecided conflict
61. Insert attack into
unresolvedUntrueList
62. Send an unresolvedUntrue(evte)
to a Voff
63. end if
64. if (sum>0 or sum<0)
65. forwardMsgtoRSUs(untrue_
dec)
66. end if
67. Clear the vector<feedback> on
untrue_id
68. end if
69. end if
70. end case
71. end while
Sender(s) and reporter(s) involved in the dispute are not permitted to participate in this
clarification process. It is reasonable to consider that there are some trusted vehicles around the
event. Also, there may be several malicious vehicles, as considered in [2, 13]. The effect of
malicious feedback will be nullified when the true feedback outweighs the malicious feedback
while taking a decision. In this framework, feedback can only be generated by trusted clarifiers
with trust scores greater than 0.5 and official vehicles. The possible feedback messages are ‘YES’
or ‘NO’. Eligible vehicles that respond, known as clarifiers, reply ‘YES’ if they had visited the
event location recently and confirmed the event or ‘NO’ if they had visited the event location and
did not see the event. In some cases, drivers neither notice the event nor visit the event location in
the recent past. These drivers will simply ignore the RSU query. Also, official vehicle feedback is
treated as the decider for a dispute which bypasses the collaboration process since collected
feedbacks from the trusted clarifiers are not used in forming a decision. When an RSU receives
official vehicle feedback in Algorithm 3, it instantly invokes the reward-punishment generator as
shown in Algorithm 4.
Algorithm 4: rew-pun-gen (rewV, punV)
Input: rewarded vehicle, punished vehicle
Output: send reward/punishment message and blocking message to TA, if required
1. while running
2. reward/punishment: //estimate
reward or punishment for disputed
event
3. Store reward(rewV) and
punishment(punV) in rewardList and
punishmentList
4. if (rewV!=Voff)
5. Send the reward_msg(rewV)
6. end if
7. Send the punishment_msg (punV)
8. Call forwardMsgtoTA(untrue_dec)
9. end while
The RSU dispute resolution mechanism in Algorithm 4 uses these feedbacks to decide the
truthfulness of a dispute. Here, the RSU performs a sum of product calculation of the feedback
and trust of the clarifiers to decide on the disputed event. For example, suppose a vector of
feedback is (‘YES’, ‘YES, ‘NO’, ‘NO’, ‘YES’) which are represented programmatically as (1, 1,
-1, -1, 1) and the clarifier’s corresponding trust scores are: (0.5, 0.7, 0.65, 0.68, 0.9), then the
RSU decides by using Eqn (3). It should be noted that only trusted clarifiers can join the
collaboration process. Generally, Eqn (3) can be expressed as in Eqn (4) for n feedbacks collected
from n trusted clarifiers, where Fi is the ith
feedback and Ti is the ith
clarifier’s trust score.
Decision = [1*0.5] + [1*0.7] + [-1*0.65] + [-1*0.68] + [1*0.9] (3)
Decision (4)

28
If the outcome is positive, then the RSU decides the sender has disseminated a true event and
thus receives an RSU reward; the conflicting reporter(s) receive an RSU punishment. If the
outcome is negative, the converse actions are followed. When the decision is reached, the RSU
calls the reward-punishment generator, shown in Algorithm 4. During the punishment
assessment, an incremental punishment policy (IPP) is applied to influence the future good
behaviour of drivers. However, for an unresolved issue when RSU has no feedback data or
Decision=0 in Eqn. (4), the RSU stores them in an unresolved dispute list and later may ask an
official vehicle to inspect the event location physically and report its findings so that the RSU can
take an action on the dispute. It should be noted that if during the collaboration process, any
official vehicle receives an RSU message, but they have not visited the disputed event location
recently, then they reply with a far-from-event message. However, if the RSU receives a decisive
message from an official vehicle, then it always decides on the event using this message and
bypasses the collaboration mechanism.
4. IMPLEMENTATION
The framework is implemented in Veins 5.0 [29] which is a tightly coupled framework for
simulating VANETs comprising the SUMO traffic simulator [30] and OMNeT++ discrete event
network simulator [31]. We extend Veins in several respects. First, we create four types of
vehicles, namely: “official” vehicles (police, ambulance, and fire service) and regular vehicles. A
TA module has been created that registers and blocks drivers. Additionally, the TA unit keeps a
driver’s most recent reward/punishment history in a driver profile database to facilitate the
blocking of malicious drivers and to record incident information (location, timestamp, incident)
in an incident database. In addition to this, an RSU internetwork is developed which also
connects to the TA unit via wired communication. Inside each RSU, besides event management,
a dispute resolution process is implemented to detect untrue/inconsistent attacks.
4.1. System Model and Environment
As there are four types of vehicles, we have created four distinct modules in OMNeT++ and the
C++ implementation of them according to the functionality specified in Section 3. A TPD
module is added to regular vehicles which primarily implements the trust update and access
blocking of each driver. The module can exchange messages with the vehicle application layer
using an internal connection. When drivers broadcast messages, they send the delay, and the
location of the event from the current location of the vehicle identified on the map. Also, the TPD
can check the current location of the vehicle and then determine the amount of
reward/punishment from an announcement using Algorithm 2. The TPD withholds rewards for a
given period to allow disputes to arise via reporters. Also, the TPD disables the transmission of a
blocked driver to stop the generation of event announcements; however, whilst in the blocking
state a driver still broadcasts beacons. The vehicle application first reads the trust from the TPD.
This trust needs to satisfy the associated trust threshold for the message class to proceed with the
announcement. The reward varies for activities like beaconing, forwarding, and broadcasting
announcements. Vehicles can only obtain a beaconing reward if they are classified as not trusted
or lowly trusted as defined in Section 3.4. The TPD can also support multiple driver profiles in
case different people share a vehicle.

29
Figure 3. Road Networks Used in the Simulation
The application layer of official vehicles is different than the regular vehicles. Official vehicles
respond to RSU queries differently than regular vehicles. Also, when an emergency arises, an
RSU gives precedence to their messages over those from regular vehicles. There is a built-in
mobility module from Veins that advances all vehicles at regular time steps. The system starts
with no vehicles in the terrain model and once a vehicle is added, it remains in the system until
the simulation terminates. Once a predetermined number of vehicles enter the system, no more
are permitted. The simulation commences by assigning periodic events to specific vehicles and
then the resultant data are collected regarding the specific experiment. The framework is
simulated using road networks from the Veins default Erlangen city map [29] as in Figure 3a, the
Manhattan grid map in Figure 3b, and one alternate route scenario in Figure 3c.
4.2. Verification – Thwarting Untrue and Inconsistent Attacks
This framework is verified in the presence of malicious and benign behaviours of trusted
vehicles. To this end, this experiment is conducted at least thirty times for 5000 simulation
seconds with 10-100 vehicles. One result from these experiments is depicted in Figure 4 to
illustrate trust management in the presence of attacks. The horizontal axis represents the
simulation time in seconds, and the vertical axis represents the trust score. All vehicles start with
a trust score of 0.9. The reward is fixed for a single announcement or RSU interaction which is
set to 0.08. and RSU punishments for three untrue announcements are set to 0.1, 0.3, and 0.5
(applying IPP) consecutively. Receivers report an event with a probability of 40% and the event
supporting probability P from clarifiers is set to 20%. Vehicle V0 constantly sends messages at
200s intervals (simulation seconds) starting from 100s. Figure 4 records the inconsistent
behaviour of V0 with the consistent behaviours of V1 and V2.

30
Figure 4. Trust Increment/Decrement with Untrue and Inconsistent Attacks
When V0 sends untrue messages, the RSU punishes it by 0.1 and 0.3 consecutively at 220s and
640s. Conversely, when V0 announces trustworthy messages consecutively at 500s, 1100s 1300s,
1500s, 1900s, the TPD adds a reserve reward at 620s, 1220s, 1420s, 1620s, and 2020s; these
rewards are withheld for 120s. V0 receives complaints for message announcements at the 700s,
900s, and 1700s. Thus, the TPD does not add any reward for these announcements. After this, the
RSU punishes V0 by 0.5 which is shown by a large reduction in the trust score in the 2020s
followed by a blocking message which sets its current trust score to 0.05 irrespective of whatever
it previously had. In this way, V0 is blocked from network access by the framework. From this
simulation result, it can be concluded that the Incremental Punishment Policy (IPP) will
demotivate vehicles from attacking repeatedly. The IPP provides a flexible means of punishing
and blocking vehicles for their inconsistent behaviours although they may sometimes announce
trustworthy messages in between their malicious activities. As vehicles receive higher
punishment in each subsequent untrue announcement, vehicles with inconsistent behaviour will
be isolated as well. Additionally, we allow only three malicious actions within the simulation
timeframe (for example 5000s) from a trusted driver to become blocked to limit further harmful
actions. So, in summary, this trace represents an example confirming the system can successfully
detect the inconsistent behaviour of a malicious vehicle and punish it accordingly.
5. PERFORMANCE EVALUATION
In this section, the evaluation of the proposed approach is investigated. The simulation model
described in Section 4 is used in the presence of varying traffic densities. Clarifiers generate
varying percentages of malicious and benevolent feedback to classify events as true negative, true
positive, false positive, and false negative observational data. Analysis shows the proposed
framework can classify events as expected. This means that when there is more benevolent
feedback the RSU can classify an event correctly and vice versa.
This set of experiments considers the generation of varying ratios of positive and negative
feedback when classifying a disputed event. Moreover, we show the minor impact of vehicle
density on the results as RSUs ignore repeated complaints regarding the same event. As expected,
when density increases, more vehicles complain about an untrue event. The RSU forwards the
first complaint to nearby RSUs which avoids invoking costly concurrent collaboration procedures
at other RSUs. Furthermore, the proposed approach is compared against a reputation approach [8]
in terms of response time and communication overhead. The response time is the decision time of

31
receiver vehicles when they receive an event message in the network. A trust management model
for a VANET can be considered the most efficient one when receivers can decide about an event
in the fastest possible time relative to trust systems where additional computation and
communication are required after the arrival of messages. Hence, response time is a good
indicator of performance. Another useful metric is the communication overhead since a trust
model with lower communication overhead reduces the burden of message transmission and
processing. For this reason, a trust model with a lower response time and communication
overhead can be regarded as superior to one where these values are higher. Furthermore, these
two metrics affect the performance of the network communication, such as channel availability,
hence the proposed approach is compared with a reputation approach which suffers from these
two factors. Results from the analysis show the proposed framework outperforms the existing one
as a receiver vehicle in the proposed framework can decide on the appropriate action without
further communication within the VANET. However, the proposed framework requires
broadcasting feedback other than the traffic event if a reporter invokes an untrue attack event.
5.1. Scenario 1 - Accuracy of the Proposed Framework
5.1.1. Simulation Setup
This series of simulations have been conducted using one predefined route for both regular and
official vehicles on the Erlangen city map from Veins [29]. Selected parameters used for
conducting the series of simulations are listed in Table 3.
We repeat each experiment five times to collect trial data for every vehicle density and
probability of supporting an event. This sample data is then averaged for analysis. We use the
probability P to control the support or denial of events from clarifiers through YES/NO
responses. For example, with a probability of P=0, clarifiers always send NO. For P=1, clarifiers
always send YES. For probabilities of P=0.2…0.8, clarifiers send YES/NO responses
accordingly. In this way, the analysis considers varying ratios of benevolent and malicious
feedback. We consider, the senders always announce true events in one set of experiments,
whereas, in another set, they always announce untrue events. One or more reporter vehicles may
send untrue attacks upon reception of these events which are also randomized with a probability
of 0.4.
Table 3. Parameters for Simulation Setup.
Selected Parameter Value
Simulation
Parameter
Simulation area 2.5km X 2.5Km
Number of vehicles [10, 30, 50, 70, 90,
100]
Number of RSUs 12
Speed of vehicles Max 80 m/s
Simulation time (seconds) 4000s
Transmission range 300m
Warm-up period 700s
Number of event source 3
Periodic announcement At 100s
Event supporting probability [0, 0.2, 0.4, 0.6, 0.8, 1]
Reward withhold timer 120s
Collaboration timer 120s
Initial trust 0.8
Attacker
Model
Untrue and Inconsistent attacks
initiated
Applies to regular
vehicles

32
The subsequent analysis considers P, the probability of being truthful or not, and D, the vehicle
density. The possible RSU judgements are given in Table 4.
Table 4. Results Classification Matrix
Predicted true Predicted false
A true event True Negative (TN) False Positive (FP)
A false event False Negative (FN) True Positive (TP)
5.1.2. Results and Analysis
In Figure 5, the x-axis represents the vehicle density, the y-axis represents the probability of
being truthful or not and the z-axis represents the normalized likelihood of classified cases. We
define the normalized likelihood of TN/FP classified cases as the ratio of the average number of
classified TN/FP cases to the average number of reported events which the RSU classifies as
TN/FP using the dispute resolution process. When the vehicle density increases, the number of
reporters also increases. However, throughout the simulations, increasing vehicle density is
shown to have only a marginal impact on the results as the RSUs ignore repeated complaints
concerning the same event from multiple reporters. This is possible as the RSU which received
the first complaint concerning an event forwards notification of it immediately to other RSUs in
the vicinity to prevent invoking further costly and redundant collaboration procedures.
Figure 5a shows the TN results for a series of simulations where sender vehicles announce only
true events. Overall, as P increases, the possibility of classifying TN cases also increases. This
means, the framework correctly classifies disputed events if most clarifiers send truthful
feedback. As expected, at P=0, there are no TN cases because all clarifiers deny the original
event. Alternatively, at P=1, all reported events are TN as all clarifiers only send YES. The TN
cases increase rapidly from P=0.4 to P=0.6 as the proportion of received YES responses is
sufficient to support the sender announcement at RSUs. Also, the number of TN cases increases
with the rise of vehicle density as the increased traffic supports the sender's announcement.
Figure 5b shows the FN results of classifying FN cases where sender vehicles announce only
untrue events which show a similar trend like TN cases. Conversely, Figure 5c shows the TP
chart for a series of simulations when sender vehicles announce only untrue messages. Overall, as
P increases, the possibility of classifying TP cases decreases. That means the approach correctly
classifies untrue events if most clarifiers rebuff the original, malicious sender message(s). The
rapid fall in TP cases noticed between P=0.4 to 0.6 arises when most clarifiers confirm the
untruthfulness of the sender’s announcement to the RSU query. Here also the number of TP cases
decreases more with increasing vehicle density as the increased traffic leads the RSU to receive
more YES responses (maliciously supporting the sender announcement). Finally, Figure 5d
shows the false FP when the sender announces only trustworthy messages which also shows a
similar trend like TP chart.

33
Figure 5. Normalized Likelihood of Classified Cases
Thus, the normalized likelihood of classified TN/FN cases is lower at lower P values and is
greater at higher P values than the expected trend. To classify an event as TN/FN, an RSU needs
more YES/NO feedback than NO/YES, respectively. The proportion of YES and NO feedback
received at an RSU is reflected in the decision which causes the curve to vary nonlinearly with P.
Also, the TP/FP curves show a nonlinear relationship for a similar reason.
5.2. Scenario 2 – Comparison with Baseline Approach
5.2.1. Simulation Setup
We implement a baseline approach [8] alongside our scheme in order to compare message flows.
We see that the receiver-side trust evaluation approach suffers badly from communication
overhead due to trust metric dissemination as receivers are busy with trust verification after the
arrival of messages. In [8] the trustworthiness of a sender is decided using one of the following
schemes: majority voting, weighted voting by reputation, and highest reputation level. The
feedback is collated at the RSU, and the trust score is subsequently interrogated. This set of
experiments runs for 800 simulation seconds and is repeated 10 times to obtain the average
number of messages exchanged in the presence of 10 to 70 vehicles. An event is introduced
deliberately at 400 seconds in both approaches. In approach [8], all vehicles upon observing the
event, announce it. Conversely, in the proposed framework the announcement of an event from
one vehicle is adequate with receivers relaying it up to 4 hops.
5.2.2. Results and Analysis
In Figure 6, the x-axis represents the number of vehicles present in the simulation, and the y-axis
represents the communication overhead for a single event. This framework is compared against
the approach in [8] with 30 and 45-second interval timers. It is clear from Figure 6 that the
overhead is higher in [8] with both timer durations than in the proposed framework. With a 30s

34
timer in approach [8], the communication overhead is two, three, and four times higher than the
proposed framework when the number of vehicles is 50, 60, and 70, respectively. In most
situations, the overhead in the proposed framework is significantly lower than the approach [8],
which suffers from a higher overhead due to the need of generating feedback towards RSU for
regular reputation updates.
Figure 6. Communication Overhead Comparison
In addition, as expected, the proposed framework is better when we compare its response time
against a receiver end-based trust approach. Approach [8] starts a timer for a predefined period
i.e., the 30s, to collect additional messages about the same event. When it expires, receivers
decide on an event. That is why it suffers from a higher response time. During this time, vehicles
may enter a “problematic” road area as they are typically moving fast. On the other hand, the
proposed framework quickly decides an event without further communication unless it is
disputed. Thus, the proposed framework exhibits a faster response time in comparison with
schemes such as [2, 8, 9, 12, 14, 20, 23].
6. CONCLUSION
In summary, this research proposes a TPD-based sender-side trust management framework for
VANETs. The framework reduces trust metric communications at the expense of equipping every
regular vehicle with a TPD. TPD installation is a one-time cost whereas, the circulation of trust
metrics is continuous assuming the communication is ongoing. Senders are trusted by default and
so response times are reduced at the receivers as trust confirmation is avoided. Additionally,
results suggest that the RSUs can successfully resolve any true/false complaints and can detect
untrue and inconsistent attacks if the majority of clarifiers send truthful feedback. Furthermore,
the framework can be enhanced with additional functionality. For example, a driver profile-based
reward/punishment database could be added where historical information concerning driver
misdemeanours can tune subsequent rewards and punishments.
ACKNOWLEDGEMENTS
This work is supported by the University of Dhaka under the Bangabandhu Overseas Scholarship
scheme funded by the People’s republic of Bangladesh government for improving quality of
education and research.

35
REFERENCES
[1] Shrikant Tangade, Sunilkumar S. Manvi, “Trust management scheme in VANET: Neighbour
communication-based approach,” in Proc. IEEE Int. Conf. on Smart Technol. for Smart Nat.
(SmartTechCon)., Bengaluru, India, 2017, pp. 741-744.
[2] Zhexiong Wei, Fei Richard Yu, Azzedine Boukerche, “Trust based security enhancements for
vehicular ad hoc networks,” in Proc. of the 4th ACM Int. Symp. on Dev. and Anal. of Intell. Veh.
Netw. and Appl. (DIVANet)., Montreal, Canada, 2014, pp. 103-109.
[3] Shrikant Tangade, Sunilkumar S. Manvi, “CBTM: Cryptography based trust management scheme for
secure vehicular communications,” in Proc. IEEE 15th Int. Conf. on Control., Autom., Robot. and
Vis. (ICARCV)., Singapore, 2018, pp. 325-330.
[4] Dahiya, Rohan, Frank Jiang, Robin Ram Doss, “A Feedback-Driven Lightweight Reputation Scheme
for IoV,” in Proc. IEEE 19th Int. Conf. on Trust. Secur. and Priv. in Comput. and Commun.
(TrustCom)., Guangzhou, China, 2020, pp. 1060-1068.
[5] Farhan Ahmad, Virginia NL Franqueira, Asma Adnane, “TEAM: A trust evaluation and management
framework in context-enabled vehicular ad-hoc networks,” IEEE Access, vol. 6, pp. 28643-28660,
May. 2018.
[6] Tahani Gazdar, Abdelfettah Belghith, Hassan Abutair, “An enhanced distributed trust computing
protocol for VANETs,” IEEE Access, vol. 6, pp. 380-392, Oct. 2017.
[7] Nadia Haddadou, Abderrezak Rachedi, Yacine Ghamri-Doudane, “A job market signaling scheme for
incentive and trust management in vehicular ad hoc networks,” IEEE trans. on Veh. Technol., vol. 64,
no. 8, pp.3657-3674. Sept. 2014.
[8] Ricardo Mühlbauer, João Henrique Kleinschmidt, “Bring your own reputation: A feasible trust
system for vehicular ad hoc networks,” J. of Sens. and Actuator Netw., vol. 7, no. 3, p.37, Sept. 2018.
[9] Siri Guleng, Celimuge Wu, Xianfu Chen, Xiaoyan Wang, Tsotumu Yoshinaga, Yusheng Ji,
“Decentralized Trust Evaluation in Vehicular Internet of Things,” IEEE Access, vol. 7, pp. 15980-
15988, Jan. 2019.
[10] Zhaojun Lu, Gang Qu, Zhenglin Liu, “A survey on recent advances in vehicular network security,
trust, and privacy,” IEEE Trans. on Intell. Transp. Syst., vol. 20, no. 2, pp.760-776, Apr. 2018.
[11] Yu-Chih Wei, Yi-Ming Chen, “Adaptive decision making for improving trust establishment in
vanet,” in Proc IEEE 16th Asia-Pacific Netw. Oper. and Manag. Symp. (APNOMS)., Hsinchu,
Taiwan, 2014, pp. 1-4.
[12] Wenjia Li, Houbing Song, “ART: An attack-resistant trust management scheme for securing
vehicular ad hoc networks,” IEEE Trans. on Intell. Transp. Syst., vol. 17, no. 4, pp. 960-969, Nov.
2015.
[13] Zhe Yang, Kan Yang, Lei Lei, Kan Zheng, Victor C.M. Leung, “Blockchain-based decentralized trust
management in vehicular networks,” IEEE Internet of Things J., vol. 6, no. 2, pp.1495-1505, May.
2018.
[14] Seyed Ahmad Soleymani, Abdul Hanan Abdullah, Mahdi Zareei, Mohammad Hossein Anisi, Cesar
Vargas-Rosales, Muhammad Khurram Khan, Shidrokh Goudarzi , “A secure trust model based on
fuzzy logic in vehicular ad hoc networks with fog computing,” IEEE Access, vol. 5, pp. 15619-
15629, Jul. 2017.
[15] Qin Li, Amizah Malip, Keith M. Martin, Siaw-Lynn Ng, Jie Zhang,, “A reputation-based
announcement scheme for VANETs,” IEEE Trans. on Veh. Technol., vol. 61 no. 9, pp.4095-4108,
Nov. 2012.
[16] Sarah Ali Siddiqui, Adnan Mahmood, Quan Z. Sheng, Hajime Suzuki, Wei Ni, “A time-aware trust
management heuristic for the Internet of Vehicles,” In 2021 IEEE 20th Int. Conf. on Trust., Secur.
and Priv. in Comput. and Commun. (TrustCom)., Shenyang, China, Oct. 2021. pp. 1-8,
[17] Muhammad Usman Aftab, Mehdi Hussain, Anders Lindgren, Abdul Ghafoor, “Towards a distributed
ledger based verifiable trusted protocol for VANET”, in 2021 Int. Conf. on Digit. Future and Transp.
Technol. (ICoDT2),. Islamabad, Pakistan, May. 2021, pp. 1-6 .
[18] Chaker Abdelaziz Kerrache, Nasreddine Lagraa, Carlos T. Calafate, Abderrahmane Lakas, “TFDD:
A trust-based framework for reliable data delivery and DoS defense in VANETs,” Veh.
Commun., vol. 9, pp. 254-267, Dec. 2017.
[19] Sashi Gurung, Dan Lin, Anna Squicciarini , Elisa Bertino , “Information-oriented trustworthiness
evaluation in vehicular ad-hoc networks,” in Proc. Netw. and Syst. Secur. (NSS)., Berlin Heidelberg,
Germany, Springer, 2013, pp. 94-108.

36
[20] Danda B. Rawat, Gongjun Yan, Bhed B. Bista, Michele C. Weigle, “Trust on the Security of Wireless
Vehicular Ad-hoc Networking,” Ad Hoc Sens. Wirel. Netw., vol. 24, no. 3/4, pp. 283-305, Jan. 2015.
[21] Guanghao Wang, Yue Wu, “BIBRM: A bayesian inference-based road message trust model in
vehicular ad hoc networks,” in Proc 13th Int. Conf. on Trust. Secur. and Priv. in Comput. and
Commun. (TrustCom)., Beijing, China, 2014, pp. 481-486.
[22] Karim Rostamzadeh, Hasen Nicanfar, Narjes Torabi, Sathish Gopalakrishnan, Victor C. M. Leung,
“A context-aware trust-based information dissemination framework for vehicular networks,” IEEE
Internet of Things J., vol. 2, no. 2, pp.121-132, Apr. 2015.
[23] Ibrahim Abdo Rai, Riaz Ahmed Shaikh, Syed Raheel Hassan, “A hybrid dual-mode trust
management scheme for vehicular networks,” Int. J. of Distrib. Sens. Netw., vol. 16 no. 7,
pp.1550147720939372, Jul. 2020.
[24] Yao Xuanxia, Zhang Xinlei, Ning Huansheng, Li Pengjian, “Using trust model to ensure reliable data
acquisition in VANETs,” Ad Hoc Netw., vol. 55, pp.107-118, Feb. 2017.
[25] Rasha Jamal Atwa, Paola Flocchini, Amiya Nayak, “Risk-based trust evaluation model for
VANETs,” In 2020 Int. Symp. on Netw., Comput. And Commun. (ISNCC)., Montreal, QC,
Canada, Oct. 2020, pp. 1-6.
[26] Honghao Gao, Can Liu, Yuyu Yin, Yueshen Xu, Yu Li, “A hybrid approach to trust node assessment
and management for vanets cooperative data communication: Historical interaction
perspective,” IEEE Trans. on Intell. Transp. Sys. pp . vol. 23, no. 9, pp. 1-10, Nov. 2021.
[27] Zhigang Yang, Ruyan Wang, Dapeng Wu, Boran Yang, Puning Zhang, “Blockchain-enabled Trust
Management Model for the Internet of Vehicles,” IEEE Internet of Things J., Oct. 2021.
[28] Xiao Chen, Jie Ding, Zhenyu Lu, “A decentralized trust management system for intelligent
transportation environments,” IEEE Trans. on Int. Transp. Sys., vol. 23, no. 1, pp.558-571, Jan. 2022.
[29] Car2x, Vehicles in Network Simulation (Veins) v5.0 (Version 5.0), car2x, September 19, 2022.
https://veins.car2x.org/download/.
[30] OpenSim Limited, Objective Modular Network Testbed in C++ (OMNeT++) v5.5.1 (Version 5.5.1),
omnetpp.org, September 19, 2022. https://omnetpp.org/software/2019/05/31/omnet-5-5-released.
[31] The German Aerospace Center, Simulation of Urban Mobility (SUMO) v1.2.0. (Version 1.2.0),
sourceforge, September 19, 2022. https://sourceforge.net/projects/sumo/files/sumo/version%201.2.0/.
AUTHORS
Rezvi Shahariar received his B.Sc. degree in Computer Science from the University of
Dhaka, Bangladesh in 2006; and an M.S. degree in Computer Science in 2007 from the
same institution. After some time as a lecturer at the University of Asia Pacific, Dhaka,
Bangladesh, he is now an Assistant Professor at the Institute of Information Technology,
University of Dhaka, whilst pursuing a PhD at Queen Mary University of London. His
research interests include wireless network analysis with an emphasis on trust, security in
VANETs, and the application of machine learning to security.
Chris Phillips (MIEEE) received a BEng. Degree in Telecoms Engineering from Queen
Mary, University of London (QMUL) in 1987 followed by a PhD on concurrent discrete
event-driven simulation, also from QMUL. He then worked in industry as a hardware and
systems engineer with Bell Northern Research, Siemens Roke Manor Research and Nortel
Networks, focusing on broadband network protocols, resource management and resilience.
In 2000 he returned to QMUL as a Reader. His research focuses on management
mechanisms to enable limited resources to be used effectively in uncertain environments.

A Trust Management Framework for Vehicular Ad Hoc Networks

More Related Content

Similar to A Trust Management Framework for Vehicular Ad Hoc Networks (20)

More from ClaraZara1 (20)

Recently uploaded (20)

A Trust Management Framework for Vehicular Ad Hoc Networks