Machine Learning and Reputation Based Misbehavior Detection in Vehicular Communication Networks

Vehicular networks are vulnerable to various attacks such as Sybil, denial-of-service (DoS) and false alert generation attacks. Cryptographic methods can provide some protection to vehicular networks from external attacks but are found to be vulnerable to internal attacks. A misbehavior detection system (MDS) can be deployed to detect and prevent internal attacks. In this paper, we propose a machine learning and reputation based MDS to enhance the detection accuracy as well as to ensure the reliability of both vehicles and messages. Proposed MDS is trained using datasets generated through extensive simulation based on the realistic vehicular network environment. To improve the accuracy of the detection, we have employed the Dempster-Shafer (DS) theory-based collaborative misbehavior detection system. In the proposed scheme, the reputation score of each vehicle is used as a belief value for Dempster-Shafer based feedback combination. In addition, we propose a beta distribution based reputation update and revocation scheme. Moreover, we show that our proposed scheme is better compared to previous methods in terms of accurately identifying various misbehaviors.