Distributed Deception Attack Detection in Platoon-Based Connected Vehicle Systems

This article investigates the distributed sensor deception attack and estimation for a class of platoon-based connected vehicles. In these systems, the information of each vehicle's position and its relative distances with respect to its neighbours plays a crucial role to achieve the desired group performance. When getting deception attacks on such important information, it will result in severe consequences including collision. In order to detect and estimate deception attacks, a longitudinal dynamic model of vehicle platoons with modelling uncertainties, measurement noises and piece-wise constant deception attacks on sensors is first presented. With the consideration of the practical issue that a local vehicle is not able to access to global information, a distributed Kalman filter is proposed to estimate the state information using local output information. Based on the residuals obtained by the distributed Kalman filter, a modified generalized likelihood ratio (GLR) algorithm is proposed to detect and estimate the sensor deception attacks. Finally, simulations with the help of standard Carsim software are provided to verify the effectiveness of the proposed algorithm. Fair comparisons between the proposed method and the standard $\chi ^2$ detector are also presented in order to provide insights for engineering practitioners.