Kalman Filter based Attack Detection for Adaptive Cruise Control Systems

In this paper, we investigate the security issues of adaptive cruise control (ACC) systems for autonomous vehicles. Different from the research on cyber-layer security, we focus on cyber-physical system (CPS) security which incorporates the vehicle's control and dynamics into the security design. More specifically, we first analyze the longitudinal dynamics of on-road vehicles and then formulate a leader-follower model for the ACC system. Moreover, we establish a continuous-time Kalman filter for the leader-follower model and characterize its innovation's probabilistic properties. Finally, we develop multiple attack detectors based on Kalman filtering and statistical testing. The effectiveness of the proposed methods is validated via numerical simulations and experiments on Carla, an open-source platform for autonomous driving research.