Vehicular ad hoc networks (VANET) is a class of ad hoc networks build to ensure the safety of traffic. This is important because accidents claim many lives. Trust and security remain a major concern in VANET since a simple mistake can have catastrophic consequence. A crucial point in VANET is how to trust the information transmitted when the neighboring vehicles are rapidly changing and moving in and out of range. Current reputation systems for VANET try to establish trust between entities, which might not be required for practical scenarios. Due to the ephemeral nature of VANET, reputation schemes for mobile ad hoc networks (MANETs) cannot be applied to VANET. In this thesis, we point out several limitations of reputation trust management schemes for VANET. In particular, we identify the problem of information cascading and oversampling, which commonly arise in social networks. Oversampling is a situation in which a node observing two or more nodes, takes into consideration both their opinions equally without knowing that they might have influenced each other in decision making. We show that simple voting for decision making, leads to oversampling. We propose a solution to overcome this problem in VANET. We also suggest new ways to merge reputation schemes with misbehavior detection schemes to establish a trustworthy VANET.

Vehicular ad hoc networks (VANET) is a class of ad hoc networks build to ensure the safety of traffic. This is important because accidents claim many lives. Trust and security remain a major concern in VANET since a simple mistake can have catastrophic consequence. A crucial point in VANET is how to trust the information transmitted when the neighboring vehicles are rapidly changing and moving in and out of range. Current reputation systems for VANET try to establish trust between entities, which might not be required for practical scenarios. Due to the ephemeral nature of VANET, reputation schemes for mobile ad hoc networks (MANETs) cannot be applied to VANET. In this thesis, we point out several limitations of reputation trust management schemes for VANET. In particular, we identify the problem of information cascading and oversampling, which commonly arise in social networks. Oversampling is a situation in which a node observing two or more nodes, takes into consideration both their opinions equally without knowing that they might have influenced each other in decision making. We show that simple voting for decision making, leads to oversampling. We propose a solution to overcome this problem in VANET. We also suggest new ways to merge reputation schemes with misbehavior detection schemes to establish a trustworthy VANET.

Vehicular ad hoc networks (VANET) is a class of ad hoc networks build to ensure the safety of traffic. This is important because accidents claim many lives. Trust and security remain a major concern in VANET since a simple mistake can have catastrophic consequence. A crucial point in VANET is how to trust the information transmitted when the neighboring vehicles are rapidly changing and moving in and out of range. Current reputation systems for VANET try to establish trust between entities, which might not be required for practical scenarios. Due to the ephemeral nature of VANET, reputation schemes for mobile ad hoc networks (MANETs) cannot be applied to VANET. In this thesis, we point out several limitations of reputation trust management schemes for VANET. In particular, we identify the problem of information cascading and oversampling, which commonly arise in social networks. Oversampling is a situation in which a node observing two or more nodes, takes into consideration both their opinions equally without knowing that they might have influenced each other in decision making. We show that simple voting for decision making, leads to oversampling. We propose a solution to overcome this problem in VANET. We also suggest new ways to merge reputation schemes with misbehavior detection schemes to establish a trustworthy VANET.

Vehicular ad hoc networks (VANET) is a class of ad hoc networks build to ensure the safety of traffic. This is important because accidents claim many lives. Trust and security remain a major concern in VANET since a simple mistake can have catastrophic consequence. A crucial point in VANET is how to trust the information transmitted when the neighboring vehicles are rapidly changing and moving in and out of range. Current reputation systems for VANET try to establish trust between entities, which might not be required for practical scenarios. Due to the ephemeral nature of VANET, reputation schemes for mobile ad hoc networks (MANETs) cannot be applied to VANET. In this thesis, we point out several limitations of reputation trust management schemes for VANET. In particular, we identify the problem of information cascading and oversampling, which commonly arise in social networks. Oversampling is a situation in which a node observing two or more nodes, takes into consideration both their opinions equally without knowing that they might have influenced each other in decision making. We show that simple voting for decision making, leads to oversampling. We propose a solution to overcome this problem in VANET. We also suggest new ways to merge reputation schemes with misbehavior detection schemes to establish a trustworthy VANET.

The safety and efficiency of vehicular communications rely on the correctness of the data exchanged between vehicles. Location spoofing is a proven and powerful attack against Vehicle-to-everything (V2X) communication systems that can cause traffic congestion and other safety hazards. Recent work also demonstrates practical spoofing attacks that can confuse intelligent transportation systems at road intersections. In this work, we propose two self-reliant schemes at the application layer and the physical layer to detect such misbehaviors. These schemes can be run independently by each vehicle and do not rely on the assumption that the majority of vehicles is honest. We first propose a scheme that uses application-layer plausibility checks as a feature vector for machine learning models. Our results show that this scheme improves the precision of the plausibility checks by over 20% by using them as feature vectors in KNN and SVM classifiers. We also show how to classify different types of known misbehaviors, once they are detected. We then propose three novel physical layer plausibility checks that leverage the received signal strength indicator (RSSI) of basic safety messages (BSMs). These plausibility checks have multi-step mechanisms to improve not only the detection rate, but also to decrease false positives. We comprehensively evaluate the performance of these plausibility checks using the VeReMi dataset (which we enhance along the way) for several types of attacks. We show that the best performing physical layer plausibility check among the three considered achieves an overall detection rate of 83.73% and a precision of 95.91%. The proposed application-layer and physical-layer plausibility checks provide a promising framework toward the deployment of on self-reliant misbehavior detection systems.

Les réseaux véhiculaires sont constitués de véhicules capables de s’échanger des informations par voie radio afin d'améliorer la sécurité routière (diffusion de messages d'alerte en cas d’accident ou de ralentissement anormal, conduite collaborative entre véhicules…) ou de permettre aux passager d’accéder à l’Internet (applications de réseaux collaboratifs, jeux interactifs, gestion des espaces libres dans les parkings…). Malheureusement, les messages liés à la sécurité routière échangés entre les véhicules peuvent être falsifiés ou éliminés par des entités malveillantes afin de causer des accidents et mettre en péril la vie des personnes. Dans cette thèse, nous nous concentrons particulièrement sur la définition, conception et l’évaluation d’une solution de sécurité pour les communications entre véhicules afin d’assurer une communication sécurisée et un bon niveau de confiance entre les différents véhicules participants. En adoptant un modèle basé sur la formation de groupes, nous procédons à l'évaluation de niveau de confiance des véhicules participants à ces réseaux et nous développons un modèle de confiance qui sert à analyser leurs comportements dans leurs groupes respectifs tout en respectant la vie privée des participants et en maintenant une surcharge minimale dans le réseau. Ensuite, nous proposons un modèle hiérarchique et modulaire permettant la détection de comportement malveillant et la gestion de la révocation des certificats des véhicules concernés / VANETs (Vehicular Ad-hoc Networks) consist of vehicles capable of exchanging information by radio to improve road safety (alerts in case of accidents or in case of abnormal slowdowns, collaborative driving…) or allow internet access for passengers (collaborative networks, infotainment, etc.). Road safety messages exchanged between vehicles may be falsified or eliminated by malicious entities in order to cause accidents and endanger people life. In this thesis, we focus on defining, designing and evaluating a security solution for V2V communications in VANET, to ensure a secure communication and a good level of confidence between the different participating vehicles. Adopting a group-based model, we consider the Trustworthiness evaluation of vehicles participating in VANET and we develop a Trust Model to analyze the behavior of the vehicles in the group while preserving the privacy of the participants and maintaining low network overhead. We then propose a hierarchical and modular framework for Misbehavior Detection and Revocation Management