A Cyber-Attack Detection in Vehicle-to-Infrastructure Communication Based on LSTM Network

This paper proposes a method to detect the cyber-attacks by cross checking Signal Phase and Timing (SPaT) information and connected vehicle trajectories data. The proposed V2I cyber-attacks detection algorithm includes the following features: i) effective in detecting the authenticity of broad-casted SPaT information; ii) feasible to identify the occurrence of cyber-attacks when the protection function fails; iii) functional in the environment of partially connected and automated traffic; iv) applicable for isolated signalized intersection. Furthermore, a sensitivity analysis on CV penetration rate is conducted. The best performance is observed when the penetration rate is 100%. Under this penetration rate, the experimental result shows that when the sample data size is greater than 350, the accuracy rate of the proposed method reached 94%.