RICAV: RIsk based Context-Aware Security Solution for the
Intra-Electric Vehicle Network
Yosra Fraiji
1,2
, Lamia ben Azzouz
1
, Wassim Trojet
2
, Ghaleb Hoblos
2
and Leila Azouz Saidane
1
1
RAMSIS Team, CRISTAL Laboratory, National School of Computer Science, 2010 Campus University, Manouba, Tunisia
2
Normandie Univ., UNIROUEN, ESIGELEC, IRSEEM, 76000 Rouen, France
Keywords: Context-Aware Security Solution, Risk, Intra-Vehicle Network, Electric Vehicle, Game Theory.
Abstract: Smart electric vehicles are equipped with many ECU (Electronic Control Unit) that provide high levels of
safety and comfort to the drivers. However, the intra-vehicle networks are targeted by hackers as they are of
great interest both in terms of processing power (botnets) and in terms of economic value (ransomware).
Therefore, static security solutions were proposed, both by researchers and car manufacturers, to secure the
Intra-Electric Vehicle Sensors network (IVSN). However, these solutions are energy-intensive and could
deplete the battery along the travel, affecting the driver safety.For this purpose, we aim to propose an adaptive
security solution, called RIsk-based Context-Aware security solution for the intra-Vehicle network (RICAV),
that considers the electric vehicle context (energy, distance to the charging stations, traffic state, etc) and the
risk assessment value to provide a trade-off between security and energy consumption. Simulation experi-
ments were conducted to evaluate the proposed approach in terms of robustness and energy consumption.
1 INTRODUCTION
Electric vehicles have taken a great attention from
government and car manufacturers in order to
improve environment wellbeing. However, the
adoption of wireless technologies for the intra-vehicle
communication has raised more security concerns.
Many attacks can be performed on the intra-vehicle
network such as eavesdropping, spoofing, DoS, etc
(Reinhard et al., 2020). Authors in (Nie, Liu, and Du
2017; Pan et al., 2017) showed, the way to hack the
car functions such as the engine management
software, door locking and starting system). To avoid
cyber-attacks on the intra electric vehicle network,
existing security solutions implement the most robust
security mechanisms (Corbett et al., 2018). In (Fraiji
et al., 2019), authors showed that according to the
cryptographic algorithm (AES, 3DES, etc)
implemented, the energy consumption could increase
by about 15% of the battery capacity. Existing
connected vehicles security solutions were designed
for carbon cars and use permanently the most robust
security mechanisms. For EV, that could deplete the
battery along the travel and could affect
the driver
safety.
Hence, static solutions are not suitable for the
electric vehicle ecosystem. Therefore, (Fraiji et al.,
2019), authors proposed a Context-Aware SecurIty
solution for the Electric Vehicle (CASIEV) that
provides a trade-off between security and energy
consumption. In this solution, the context of the
vehicle (energy, distance to the charging station,
traffic, etc) is considered, when the battery level is
critical, to secure the system as long as possible along
the route.
In another hand, the risk is defined as Threat
likelihood × Impact (NIST, 2012). The likelihood
estimates the attack feasibility (probability of
success). The Impact (also called severity) indicates
the assessment of the risk level and intensity. Many
works (ETSI, 2017), (Kaveh Bakhsh Kelarestaghi,
Mahsa Foruhandeh, Kevin Heaslip, 2019), (Shaikh
and Thayananthan, 2019), in the literature,
investigated the risk assessment. Furth-ermore, some
works designed security solutions based on the risk
(Arfaoui et al., 2018; Gebrie and Abie 2017; Pham,
Makhoul, and Saadi, 2011), (Atlam et al., 2020).
Indeed, the authors propose approaches adapting the
level of security to network intrusion risk value.
The main concern in this work is to combine the
context used in CASIEV (focusing mainly on
enhancing the energy delivery process) with the risk
assessment of an intrusion into the intra-EV network
in order to improve real-time decision on the relevant
772
Fraiji, Y., ben Azzouz, L., Trojet, W., Hoblos, G. and Saidane, L.
RICAV: RIsk based Context-Aware Security Solution for the Intra-Electric Vehicle Network.
DOI: 10.5220/0010581407720778
In Proceedings of the 18th Inter national Conference on Security and Cryptography (SECRYPT 2021), pages 772-778
ISBN: 978-989-758-524-1
Copyright
c
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
security level activation. The new approach we called
RICAV RIsk-based Context-Aware security solution
for the intra-Vehicle network improves CASIEV by
extending more the lifetime of the security system.
Indeed, according to RICAV, there is no need for a
high level of security when the risk is low, even if
there are no energy constraints. However, according
to CASIEV if there are no energy constraints, the
security level must be high. RICAV is modelled using
game theory considering two players (energy
management system and security system) as game
theory is suitable for modelling systems with
conflicting objectives in order to find the trade-off
between them. The rest of this paper is organized as
follows: Section 2 presents a risk assessment
background. Section 3 describes the RICAV system.
Simulation results are discussed in Section 4. A brief
conclusion addresses the contributions and
perspectives of this work.
2 RISK ASSESSMENT
BACKGROUND
In the literature, the risk assessment has attracted the
attention of researchers and standardization bodies
that issued several standards. The NIST (USA
National Institute of Standards and Technology) risk
assessment (NIST, 2012) includes system
characterization, threat sources and events, system
vulnerabilities identification, security countermea-
sures evaluation and risk determination (impact-
likelihood matrix). Therefore, it detects, evaluates
and prioritizes risks. The ETSI TVRA (Threat,
Vulnerability and Risk Analysis) study the risk in the
context of the vehicular network. It identifies risks,
their likelihood and impact. Furthermore, it involves
seven steps: identify security objectives and security
requirements, produce an inventory of system assets,
classify system vulnerabilities and threats, quantify
the likelihood and impact of attacks, determine the
risks involved and specify detailed security
requirements. The SecRAM (Marotta et al., 2013)
method is the ISO 27005 based risk assessment
management methodology that was developed for air
traffic management. It associates a value between 1
and 5 to the threat impact on the security services
(Availability (Av), Authentication (Au),
Confidentiality (C), Integrity (I) and Non-repudiation
(Nr)). Furthermore, it considers the highest impact
service (Av, Au, C, I and Nr) as an overall threat
impact. Many works in the literature investigated risk
assessment in the context of the in-vehicle network.
In (Kaveh Bakhsh Kelarestaghi, Mahsa Foruhandeh,
Kevin Heaslip, 2019), authors adapted (NIST, 2012)
in the context of a compromised in-vehicle network.
The goal of this methodology is to explore threats
targeting the in-vehicle networks and to map impacts
of such threats into risk clusters. For example, they
consider safety and behavioural impacts as a very
high risk. In (Shaikh and Thayananthan, 2019)
authors proposed a fuzzy risk-based decision for
vehicular networks while adopting the NIST risk
definition at a high level. However, they use new
mechanisms to identify the likelihood and impact
value. The likelihood is based on the vehicle context
(lane, road, traffic, weather, speed and time) and the
driver’s attitude. On the other hand, the impact is
calculated based on the type of application. The
EVITA project(Ruddle and Ward, 2009) (Esafety
Vehicle Intrusion Protected Applications) proposed
an intra-vehicle risk assessment. It is considered as a
prominent risk assessment model that adopted the
asset-oriented approach. EVITA proposed a risk
matrix that includes attacks likelihood, the attack
severity, and the driver controllability. The severity is
calculated in terms of four factors: Safety, privacy of
drivers, operational performance, and financial
losses. The likelihood of a threat is considered in
terms of the expertise, knowledge of target, window
of opportunity (including time requirement).
3 RISK-BASED
CONTEXT-AWARE SECURITY
SOLUTION FOR THE
INTRA-VEHICLE NETWORK
The Intra-Vehicle Network is a complex network of
sensors, ECUs and firmware that can be vulnerable to
many types of attacks. Our goal is to minimize the
energy consumption of the system while maintaining
the security of the Intra-Vehicle Network
communications as long as possible along the route.
3.1 RICAV Architecture
Figure 1 presents the architecture of RICAV. It is
composed of two systems: CASIEV (Context-Aware
Security for the Intra-Electric Vehicle) (Fraiji et al.,
2019) and ASR (Adaptive Security based on the
Risk). CASIEV adapts the security of the in-vehicle
sensors network according to the EV dynamic
context. We defined the context as the State of Charge
(SoC), the nearest available charging station, the
sensor resources (memory and processing) and the
RICAV: RIsk based Context-Aware Security Solution for the Intra-Electric Vehicle Network
773
traffic conditions. CASIEV applies the high security
level allowed by the context without considering the
risk probability. The ASR module chooses the
required security level according to the risk in order
to improve the energy saving process. If the risk is
low, there is no need to ask for a high security level
which is an energy incentive.
Figure 1: RICAV architecture.
3.2 Which Modelling Approach for
RICAV?
In this work, we face a multi-objective optimization
problem as it requires more than one objective
function to be optimized simultaneously (preserving
security and optimizing energy).In the literature,
many techniques are used to solve this problem such
as Weighted-sum method, e-constraints Method,
Multi-level Programming, Goal Programming,
Evolutionary Algorithm (Genetic Algorithm,
Differential Evolution) and game theory. In (Sfar et
al., 2019), authors showed that game theory
outperforms many well-known multi-objective and
meta-heuristic algorithms in quality, stability,
convergence speed, and running time.Game theory
balances the trade-off between conflicting objectives
(Liang and Xiao, 2013).Furthermore, it is adapted to
this case study as it can model scenarios in which
there is no centralized entity with a full picture of
network conditions.Indeed, the security system does
not have any information about the battery SoC (State
of Charge) and vice versa.
Game theory is used in network security as a
quantitative framework which studies the interaction
between hackers and defenders(Liang and Xiao,
2013). In fact, many authors adopted the Game
theoretic approach to model adaptive security
solutions. In (Hamdi and Abie, 2014), authors
proposed a game theoretical model for an e-Health
adaptive security preserving the authentication of
smart things. Authors provided a mathematical
model, relying on Markov game theory, to present
healthcare under dynamic context. This model, based
on a set of strategies to design the game model, uses
four basic parameters to represent the context
(memory, communication channel, energy depletion
model and the threat model). In (Xiaolin et al. 2008),
authors developed an adaptive security model relying
on Markov chain for the network information system.
This work is based on two Markov chains. The first
chain was used to model the propagation of both the
threats in the network and the quantified risk. The
second one adapted the security of the system
according to the quantified risk.
3.3 Game based Risk and
Context-Aware Security
Formulation
RICAV (RIsk based Context Aware Security for the
intra-Vehicle network) is a game-based risk and
context aware solution. In the considered scenario,
players compete for the limited network resources (in
our case: energy). In this section, we will present the
parameters, assumptions, game specification, game
tree, Nash equilibrium and the behavioural System
Model.
3.3.1 Assumptions
We assume that:
The proposed game is a Non-Cooperative Dynamic
Game Incomplete Information in which two players
compete with each other. A non-cooperative game
is one in which any cooperation must be self-
enforcing, as players are strictly rational and play to
optimize their individual expected value. The game
is dynamic as we consider a dynamic vehicular
context (dynamic energetic context and risk). In the
complete game all players have the same privileges
and knowledge about the game conditions and the
other players’ strategies and actions.In the
incomplete information players don’t have the
information of the other players, however it needs
to identify other players choices and hence predict
its behaviour.
The game is a sequential game in which players
alternate turns. The security system will play its
strategy and the energy system will in turn react.
3.3.2 Game Specification
The game 𝐺 is defined as a triplet (𝑃, 𝑆, 𝑈), where 𝑃
is the set of players, 𝑆 is the set of strategies, and 𝑈 is
the set of payoff functions (Manshaei et al., 2013). In
the proposed strategy, we consider two players: the
energy management system and the security system.
SECRYPT 2021 - 18th International Conference on Security and Cryptography
774
The energy management system represents the key
player of the game. For each player we will describe
their strategies, utility function and pay-offs. The
security system adapts the security level of sensors
according to the identified risk. The energy
management system aims to optimize the energy
consumption of the intra-electric vehicle network
according to the context. In this section, we begin by
describing the game of player 1. Then, we will
describe the game of player 2.
Players: P= {security system (ss), energy
management system (es)}.
Stage 1: Security System Player.
Let L= {𝑙
, 𝑙
,..,𝑙
} be the set of security levels𝐿
, 0
≤ i ≤ n
𝑆

is the set of strategy of the security system
𝑆

= L
Utility of the player 1:maximizing the robustness of
the security system while minimizing the overhead
(processing, memory, delay). The robustness of a
network will be assessed as the degree of the security
strategy capability to withstand attacks. The security
system adapts the security level (l) according to the
risk value 𝑟
.
𝑈

represent the Pay-offs of the security system.
U

= G (𝑝
)
The Payoff Function is modelled by a sigmoid
function, as demonstrated in (Sfar et al., 2019).The
sigmoid value is between [0, 1]. This function is
classified as a nonlinear, quickly increasing and
simple function which can meet the requirement of
calculating the required security level probability in a
reasonable running time.
the gain function G (𝒑
𝒍
)is defined as in (1)
G (𝒑
𝒍
)=
𝟏
𝟏𝒆
 𝒈
𝒍
∗ 𝒑
𝒍
𝒉
𝒍
,𝒈
𝒍
=𝒓, 𝒓 > 𝟎. 𝟎𝟓
(1)
with 𝑔
the steepness of sigmoid function,
: the
center of the sigmoid function. In
practice,𝑔
represents the risk level. We consider the
r=0 as a special case. The equation
𝟏
𝟏𝒆
 𝒈
𝒍
∗ 𝒑
𝒍
𝒉
𝒍
cannot reflect the reality as for all
probabilities value the function will return always 0.5.
For this purpose, if risk (r) is equal to zero𝑔
=10 (2).
𝑝
is the probability of using the required security
level 𝑙
.
G (𝒑
𝒍
)=
𝟏
𝟏𝒆
 𝒈
𝒍
∗ 𝒑
𝒍
, 𝒓 = 𝟎. 𝟎𝟓, 𝒈
𝒍
=𝟏𝟎
(2)
Stage 2: Energy Management System Player.
Let E= {on ,𝑜𝑓𝑓},
𝑆

be the set of strategy of Energy management
system player.𝑆

= E
The energy management system is in the on mode if
it accepts to deliver the required energy and in the off
mode otherwise.
Utility Function of the Player 2: The energy
management system provides the good operation of
the intra-vehicle network with a minimal cost
(minimizing the energy consumption).
U

represent the Pay-offs of the energy management
system (3). U

= L (𝑝
)
L (𝒑
𝒆
)=
𝟏
𝟏𝒆
𝒈
𝒆
∗( 𝒑
𝒆
𝒉
𝒆
)
(3)
with 𝑔
: the steepness of sigmoid function,
: the
center of the sigmoid function. In practice, the 𝑔
reflect the system state. It is equal to 0.05 if the
system is green, 0.5 if the system is orange, 1 if the
system is red (see table 1). 𝑝
is the probability of
delivering the required energy.
Table 1: Energy management system state.
S
y
stem
state
Description
Green state
(
𝑔
=0.05
)
The energy management system
accepts to deliver energy.
Orange state
(𝑔
=0.5)
The energy management system can
accept or refuse to deliver energy. This
decision is based on the context
parameters (charging station and
traffic
)
.
Red state
(
𝑔
=1
)
The energy management system
refuses to deliver energy.
Stage 3: General Objective Function.
The two parameters 𝑝
and 𝑝
probabilities are
defined independently. However, in the present
model the only context in which the security system
adapts the required security strategy is when it has the
required energy. We can conclude that the two events
are the same and their probabilities coincide. For this
purpose, we can define p

:= p
= p
. The objective
of the game is to maximize the function (4). It is
continuous and defined on a compact, which is easy
to prove in our case.
𝐔(𝐩
𝐥𝐫
)=(𝐆(𝐩
𝐥𝐫
)∗ (𝟏−𝐋
(𝐩
𝐥𝐫
)))
(4)
3.3.3 Equilibrium Solution
The utility functions defined above express a trade-
off between (energy and security):
Enforcing the policy (at the risk of depleting the
battery)
RICAV: RIsk based Context-Aware Security Solution for the Intra-Electric Vehicle Network
775
Using a less robust security level (at the risk of
violating the security policy).
The equilibrium of the game is denoted by (𝑒𝑞
) and
is found by solving the following optimization
problem (5).
eq: = argmax {U (𝒑
𝒍𝒓
),𝒑
𝒍𝒓
𝟎..𝟏 }
(5)
eq
is the value of 𝑝

maximizing the utility function
and thereby, reflecting the optimal probability of
disclosing the required security level. In the particular
case, we can retrieve the optimum value of eq
explicitly g
= g
and h
> h
. In the general case, the
value of eq is calculated numerically (see the
simulation section).
4 SIMULATION AND ANALYSIS
We solve the game equilibrium for different
situations in the game numerically. We represent the
gain function G (𝒑
𝒍
) and the loss function L (𝒑
𝒆
),
calculate their product, find their maximum point and
get the corresponding steady state. To simulate
different scenarios, we modify the𝑔
and 𝑔
values
(risk and energy level) and analyse the player’s
behaviour. Figures 2, 3 and 4 present results for a
scenario where the energy is available (𝑔
=0.005
green energetic state and the risk value varies. We
notice in the figures 2,3,4 a Nash equilibrium. Indeed,
both players are winners as the energy is available,
hence giving the energy may not result in an
important loss (the loss will always be moderate).
Figure 2 shows that the gain of the security system
(robustness) is very low if the probability of obtaining
the required security level is low and it can reach 1 in
the opposite case. In such scenario, the RICAV
prioritize the security then the energy consumption.
Figure 2: Nash equilibrium for risk=1 and green energetic
state.
In figure 3, since the risk is equal to 0.5, the gain
of the security system is more important than in the
previous case even for a low probability.
Figure 3: Nash equilibrium for risk=0.5 and green energetic
state.
Figure 4: Nash equilibrium for risk= 0.05 and green
energetic state.
Figure 5: Nash equilibrium for risk=0.05 and orange
energetic state.
In figure 4, we have considered a risk equal to
0.05. We notice that the robustness of the system is
high even if the security level is not provided since
the risk of attack is almost inexistent. Indeed, the
decrease in risk leads to an increase in the robustness
of the system. In this case, RICAV can ask for a low
security level (or no security at all) even though the
energy is available. Figure 5 present results for a
scenario where the energy has become critical
(𝑔
=0.5 orange energetic state) and the risk value
varies. In this case, we obtain a Nash equilibrium only
in the case where the risk is very low (r=0.05). Indeed,
SECRYPT 2021 - 18th International Conference on Security and Cryptography
776
since the battery can deliver or does not deliver
energy, there is always a winner and a loser. In this
scenario, RICAV provides a trade-off between energy
and security. It prioritizes security if the risk is high
and prioritizes energy saving if the risk is low. In the
red energetic state, we consider a battery in the red
zone where the energy becomes very critical. In this
case, we obtain a Nash equilibrium only in the case
where the risk is very low (r=0.05) since the energy
system is not allowed to supply energy in this zone.
That means, the equilibrium probability when the risk
is low (r=0.05) is not related to the decision of the
energy management system. In this scenario RICAV
prioritizes energy saving.
5 CONCLUSION
In this work, we proposed a risk-based context-aware
security solution for the intra-electric vehicle sensor
network. This solution allows the system to preserve
energy as it adapts the security according to the risk
and the vehicular context (energy, distance to
charging station, traffic, etc). RICAV is modelled
using game theory. The game is composed of two
players: the security system and the energy
management system. The security system adapts the
security level according to the identified intrusion
risk. The energy management system provides the
energy amount required by the security system
according to the vehicle context. Simulations show
that the robustness of the system grows when the risk
decreases. Therefore, RICAV prioritizes the energy
saving process if the risk is low. It prioritizes security
if the energy is available and the risk is high or
medium. For future works, we intend to improve
RICAV by developing a trust model for the intra-EV
network intrusion risk assessment based on the
vehicle current context and its previous experience.
Indeed, considering the risk trust value could enhance
the energy saving process. For example: if the risk is
high and the trust is low, the system can ask for a low
security level improving this way the energy saving
process.
REFERENCES
Arfaoui, Amel, Ali Kribeche, Sidi Mohammed Senouci,
and Mohamed Hamdi. 2018. “Game-Based Adaptive
Risk Management in Wireless Body Area Networks.”
In 2018 14th International Wireless Communications &
Mobile Computing Conference (IWCMC), IEEE, 1087–
93.
Atlam, Hany F. et al. 2020. “Risk-Based Access Control
Model: A Systematic Literature Review.” Future
Internet 12(6): 1–23.
Corbett, Christopher et al. 2018. “Leveraging Hardware
Security to Secure Connected Vehicles.” SAE
Technical Paper Series 1: 1–12.
ETSI. 2017. “Intelligent Transport Systems (ITS); Security;
Threat, Vulnerability and Risk Analysis (TVRA).” 1:
1–88.
Fraiji, Yosra et al. 2019. “Adaptive Security for the Intra-
Electric Vehicular Wireless Networks.” 2019 15th
International Wireless Communications & Mobile
Computing Conference (IWCMC): 1215–20.
Gebrie, Mattias T, and Habtamu Abie. 2017. “Risk-Based
Adaptive Authentication for Internet of Things in Smart
Home EHealth.” In Proceedings of the 11th European
Conference on Software Architecture: Companion
Proceedings, , 102–8.
Hamdi, Mohamed, and Habtamu Abie. 2014. “Game-Based
Adaptive Security in the Internet of Things for
EHealth.” 2014 IEEE International Conference on
Communications, ICC 2014: 920–25.
Kaveh Bakhsh Kelarestaghi, Mahsa Foruhandeh, Kevin
Heaslip, Ryan Gerdes. 2019. “Intelligent
Transportation System Security : Networks.”
Liang, Xiannuan, and Yang Xiao. 2013. “Game Theory for
Network Security.” IEEE Communications Surveys &
Tutorials 15(1): 472–86.
Manshaei, Mohammad Hossein et al. 2013. 45 ACM
Computing Surveys Game Theory Meets Network
Security and Privacy.
Marotta, Antonio et al. 2013. “Applying the SecRAM
Methodology in a CLOUD-Based ATM Environment.”
Proceedings - 2013 International Conference on
Availability, Reliability and Security, ARES 2013
(December): 807–13.
Nie, Sen, Ling Liu, and Yuefeng Du. 2017. “Free-Fall:
Hacking Tesla from Wireless to Can Bus.” Defcon: 1–
16.
NIST. 2012. “NIST Special Publication 800-30 Revision 1
- Guide for Conducting Risk Assessments.” NIST
Special Publication (September): 95.
Pan, L et al. 2017. “Journal of Information Security and
Applications Cyber Security Attacks to Modern
Vehicular Systems.” Journal of Information Security
and Applications 36: 90–100.
Pham, Congduc, Abdallah Makhoul, and Rachid Saadi.
2011. “Risk-Based Adaptive Scheduling in Randomly
Deployed Video Sensor Networks for Critical
Surveillance Applications.Journal of Network and
Computer Applications 34(2): 783–95.
Reinhard, Jan Peter, Marcel Kneib, Martin Ring, and Oleg
Schell. 2020. “Assessment of Current Intrusion
Detection System Concepts for Intra-Vehicle
Communication.” : 1–2.
Ruddle, Alastair, and David Ward. 2009. “Security
Requirements for Automotive On-Board Networks
Based on Dark-Side Scenarios.” (1): 138.
Sfar, Arbia Riahi, Yacine Challal, Pascal Moyal, and Enrico
Natalizio. 2019. “A Game Theoretic Approach for
RICAV: RIsk based Context-Aware Security Solution for the Intra-Electric Vehicle Network
777
Privacy Preserving Model in IoT-Based
Transportation.” IEEE Transactions on Intelligent
Transportation Systems PP: 1–10.
Shaikh, Riaz Ahmed, and Vijey Thayananthan. 2019.
“Risk-Based Decision Methods for Vehicular
Networks.” Electronics (Switzerland) 8(6).
Xiaolin, Cui, Tan Xiaobin, Zhang Yong, and Xi
Hongsheng. 2008. “A Markov Game Theory-Based
Risk Assessment Model for Network Information
System.” 2008 International Conference on Computer
Science and Software Engineering: 1057–61.
SECRYPT 2021 - 18th International Conference on Security and Cryptography
778