Physics joins AI: A real-time hybrid misbehavior detection framework for vehicular networks

The paper addresses a central requirement for cooperative driving: messages must not only arrive, they must also be physically plausible in the current vehicular context. To do that, the work combines two independent evaluations, one based on AI analysis and one based on the physical coherence of message content.

The result is a complementary approach that aims to make the final decision more reliable on both the VeReMi dataset and a highway scenario simulated with Plexe. It helps separate faults and malfunctions from attacks, and it also proposes a simple protocol for safely dismantling a platoon when suspicious messages are received.