Anomaly Detection from Cyber Threats via Infrastructure to Automated Vehicle

Résumé : Using Infrastructure-to-Vehicle (I2V) information can be of great benefit when driving autonomously in high-density traffic situations with limited visibility, since the sensing capabilities of the vehicle are enhanced by external sensors. In this research, a method is introduced to increase the vehicle's self-awareness in intersections for one of the largest foreseen challenges when using I2V communication: cyber security. The introduced anomaly detection algorithm, running on the automated vehicle, assesses the health of the I2V communication against multiple cyber security attacks. The analysis is done in a simulation environment, using cyber-attack scenarios from the Secredas Project (Cyber Security for Cross Domain Reliable Dependable Automated Systems) and provides insights into the limitations the vehicle has when facing I2V cyber attacks of different types and amplitudes and when sensor redundancy is lost. The results demonstrate that anomalies injected can be robustly detected and mitigated by the autonomous vehicle, allowing it to react more safely and comfortably and maintaining correct object tracking in intersections.