Cyber-Resilient Data-Driven Event-Triggered Secure Control for Autonomous Vehicles Under False Data Injection Attacks

This paper proposes a cyber-resilient secure control framework for autonomous vehicles (AVs) subject to false data injection (FDI) threats as actuator attacks. The framework integrates data-driven modeling, event-triggered communication, and fractional-order sliding mode control (FSMC) to enhance the resilience against adversarial interventions. A dynamic model decomposition (DMD)-based methodology is employed to extract the lateral dynamics from real-world data, eliminating the reliance on conventional mechanistic modeling. To optimize communication efficiency, an event-triggered transmission scheme is designed to reduce the redundant transmissions while ensuring system stability. Furthermore, an extended state observer (ESO) is developed for real-time estimation and mitigation of actuator attack effects. Theoretical stability analysis, conducted using Lyapunov methods and linear matrix inequality (LMI) formulations, guarantees exponential error convergence. Extensive simulations validate the proposed event-triggered secure control framework, demonstrating substantial improvements in attack mitigation, communication efficiency, and lateral tracking performance. The results show that the framework effectively counteracts actuator attacks while optimizing communication-resource utilization, making it highly suitable for safety-critical AV applications.

💡 Research Summary

This paper addresses a critical cybersecurity challenge in autonomous vehicles (AVs): false data injection (FDI) attacks targeting the steering actuator. The authors propose a novel, integrated control framework designed to be cyber-resilient, meaning it actively withstands and mitigates such adversarial interventions while maintaining system performance and communication efficiency.

The core innovation lies in the synergistic combination of three key methodologies. First, to overcome the limitations of traditional mechanistic models, a data-driven modeling approach using Dynamic Mode Decomposition (DMD) is employed. This technique extracts the essential lateral dynamics of the vehicle directly from real-world experimental sensor data, leading to a more accurate and adaptable system representation.

Second, the framework incorporates an event-triggered communication scheme. Instead of transmitting control signals at fixed time intervals, data is only sent when the system’s tracking error exceeds a predefined threshold. This significantly reduces network bandwidth usage and alleviates congestion, which is crucial for in-vehicle networks, without compromising stability.

Third, for active attack mitigation, a secure control law is developed. It integrates a Fractional-Order Sliding Mode Control (FSMC) strategy, known for its robustness against disturbances, with an Extended State Observer (ESO). The ESO is designed to estimate the actuator attack signal in real-time by treating it as an extended system state. The FSMC controller then uses this estimate to actively compensate for and neutralize the attack’s effect within the control command itself, ensuring precise path tracking even under persistent FDI attempts.

The paper provides rigorous theoretical stability analysis using Lyapunov theory and Linear Matrix Inequality (LMI) formulations. This analysis formally proves that the closed-loop system guarantees exponential convergence of the tracking error, even in the presence of bounded actuator attacks and under the event-triggered transmission protocol.

Extensive simulation studies validate the proposed framework’s effectiveness. The results demonstrate substantial improvements compared to conventional methods: superior lateral path-following accuracy under attack conditions, successful real-time attack estimation and compensation by the ESO, and a dramatic reduction in communication frequency due to the event-triggered mechanism. The proposed ETS-FSMC framework thus presents a holistic solution that simultaneously enhances security, ensures control performance, and optimizes resource utilization, making it a highly promising approach for safety-critical autonomous driving applications.