Anomaly Detection in Cooperative Adaptive Cruise Control Using Physics Laws and Data Fusion

Cooperative Adaptive Cruise Control (CACC) is a promising application in autonomous vehicles. In CACC, a platoon is formed, where a leading vehicle sends information to regulate the speed of succeeding "following" vehicles in the road. In a benign environment, CACC provides tremendous benefits, including reducing fuel consumption, maximizing road capacity, and increasing traffic safety. However, one of the critical security threats in CACC is when a platoon has a compromised leading vehicle, which forges acceleration information sent to the platoon members. Such data falsification attack would lead to traffic instability, high fuel consumption, and potential collisions. As an effort to solve this problem, in this paper, we propose a real-time anomaly detection mechanism using physics laws of kinematics along with data fusion. The proposed technique is applied at each vehicle, where the information received from the leader is validated based on physics laws. To enhance the reliability and support the detection of anomalous behavior, we utilize information sharing in CACC by allowing vehicles and the fixed infrastructure to share sensed information about platoon leaders. In the proposed approach, each vehicle fuses information it receives to reliably detect unexpected deviations. We showed that the proposed approach is effective in detecting acceleration falsification attacks in CACC, and provides high detection accuracy (96%) and negligible false alarm rate. We compare our proposed approach with existing method and showed that the proposed approach provides superior performance in both detection accuracy and execution time.