Adversarial Models for Understanding Fundamental Limitations of Attack Detection in Vehicle Platoons

Abstract Connected and Autonomous Vehicles (CAVs) are becoming a promising intelligent transportation technology which can be beneficial in reducing transportation energy consumption as well as human-induced accident risks. However, with the heavy reliance on communication and information technologies, CAVs also pose significant security risks in the form of adversarial cyber-attacks. Recent research works have been exploring various solutions to the security problems in CAVs. However, a key research gap exist in the fact that there is no standard adversarial threat model for CAV applications. Typical threat models used in analyzing security risks are mostly ad hoc. This brief attempts to address this research gap by proposing an adversarial threat model for CAVs. Specifically, we propose an adversarial threat model for vehicular platoon applications, which utilizes underlying physics model in conjunction with optimal control theory to generate intelligent attack policies. Such threat model will be useful in (i) testing the effectiveness of security solutions in a formal way, and (ii) uncovering the fundamental limitation of model and real-time data-based attack detection strategies.