A Cyberattack Warning System for Enhancing Connected Vehicle Safety Under Spoofing Cyberattacks: A Generative-Based Human-in-the-Loop Trajectory Prediction Approach

As vehicles increasingly integrate with infrastructure and each other, the risk of cyberattacks is escalating significantly. Existing research mainly focuses on the threats within the environments of connected autonomous vehicles (CAVs). However, connected vehicles (CVs) that are driven by humans are also vulnerable to spoofing attacks. The trajectory of CVs under cyberattacks has not been comprehensively explored. Besides, existing research primarily focuses on the countermeasures for detecting but lacks the strategies to be taken after a cyberattack has occurred. In this study, we propose a Cyberattack Trajectory-based Forecasting and Warning System (Cyber-TFWS) that demonstrates effective perception, prediction, judgement and warning capabilities under cyberattacks to enhance the safety against spoofing attacks in CV environment. We introduce a research framework for effectively collecting the trajectories of human-driven CVs under cyberattacks and the trajectories are processed by an unsupervised algorithm. A novel generative-based algorithm named CAGAN integrating into our system is also proposed. The results indicate the proposed system successfully issues the warning actions to the specific drivers and prevented 100% of red-light running behaviors and also significantly enhanced the safety. The proposed system has the potential to be incorporated with CV applications and can benefit governs or police markers in developing the cyberattack protection system.