Reinforcement Learning-based Misbehaviour Detection in V2X Scenarios

Emerging vehicle-to-everything (V2X) services rely on the secure exchange of periodic messages between vehicles and between vehicles and infrastructure. However, transmission of false/incorrect data by malicious vehicles may pose important security perils. Therefore, it is essential to detect safety-threatening erroneous information and mitigate potentially detrimental effects on road users. In this paper, we assess the effectiveness of a reinforcement learning (RL) approach for misbehaviour detection in V2X scenarios using an open-source dataset. Considering the case of sudden-stop attacks, the performance of RL-based detection is evaluated over commonly used detection metrics. Our research outcomes reveal that misbehaving vehicles can be accurately detected by exploiting real-time position and speed patterns.