Machine Learning to Secure Wheels: A Survey of Misbehavior Detection for the Next-Generation of Connected and Autonomous Vehicles

Connected and autonomous vehicles (CAVs) rely on 5G vehicle-to-everything (V2X) communications to enable safety critical and cooperative applications such as emergency braking, collective perception, platooning, and hazard warning. Public key infrastructure (PKI) ensures authentication and message integrity but is insufficient against insider attacks, data manipulation, and fault induced anomalies in highly dynamic vehicular environments. As a result, misbehaviour detection systems (MDSs), supported by machine learning (ML), have become an essential complement to cryptographic security. This survey presents a comprehensive and deployment-oriented review of ML-based MDSs for next-generation CAV ecosystems. Unlike existing surveys that focus mainly on attack-centric or algorithm-centric perspectives, this work explicitly considers both intentional misbehaviour, including Sybil attacks, message falsification, and denial of service, and unintentional misbehaviour arising from sensor faults, protocol deviations, and environmental interference. We introduce a novel five-dimensional classification of misbehaviour detection approaches, encompassing knowledge-driven, data-driven, trust and reputation-driven, context-aware and adaptive, and hybrid and fusion-based systems, organized around how evidential information is generated, validated, and fused. Furthermore, the survey proposes a structured taxonomy of ML-based MDSs that separates learning regimes, including supervised, unsupervised, and semi-supervised, and reinforcement learning, from deployment and adaptation mechanisms such as federated and transfer learning, reflecting the non-IID data, mobility, and privacy constraints of CAV networks. We examine representative V2X use cases to connect operational requirements with suitable detection strategies, and open challenges related to zero-day detection, adversarial robustness, real-time operation, and standardization are highlighted.