Hierarchical Game-Based Optimal Security Control of Vehicular Platoon Systems Against FDI Attacks

Zhiwei Xia, Chun Liu, Xiaoqiang Ren, Hongtian Chen, Xiaofan Wang

Abstract: This study investigates the optimal security control (OSC) problem for Intelligent Connected Vehicle Platoon Systems (ICVPS) against malicious false data injection (FDI) attacks. A hierarchical game-based control framework is proposed, which integrates differential graph games and zero-sum dynamic adversarial games. This framework is designed to guarantee both global stability and local security performance of the ICVPS. In the outer layer addressing network communication, a differential graph game model is established, treating each vehicle as an independent agent. This enables the entire ICVPS to achieve Nash equilibrium, thereby ensuring optimal global platoon stability and cooperative control precision. In the inner layer focusing on vehicle-level control, a zero-sum dynamic adversarial game model is constructed. Here, the designed optimal security controller and state information-based malicious FDI attacks act as opposing players, ensuring optimal vehicle-level security performance under dual-channel FDI attacks. Furthermore, an improved offline-learning residual neural network (OLRNN) integrated with a cosine annealing mechanism is designed to solve the optimal security controller. This approach significantly enhances training efficiency and effectively prevents the OLRNN from falling into local optima. Finally, simulation experiments conducted on the CARLA simulator verify that the proposed algorithm can effectively mitigate vehicle security issues caused by malicious FDI attacks with minimal control cost. Additionally, ablation studies confirm the superiority of the proposed OLRNN model.

External IDs:doi:10.1109/tvt.2026.3650977