A Model for Identifying and Isolating Sensor Attacks in Autonomous Vehicles

The proposed solution in this paper is model-based which aims to address the cyber-security threats affecting automated cars more so those affecting the sensors-targets. The goal of the framework is to detect the risks and find their position to provide secure positioning of the AVs. To build a tenacious protection against cyber threats the technique involves having multiple sensors to incorporate many physical sensors that give real time posture. For real-time detection of anomalies in the sensor measurements the design involves an extended Kalman filter (EKF) and a cumulative sum (CUSUM) discriminator. Iterator calculations of the position and orientation of a vehicle are carried out using Extended Kalman Filters (EKFs). At the same time, there are CUSUM discriminators employed in evaluating the differences between actual and expected positions in line with the vehicle mathematical model or failure identification. An auxiliary detector combines the information from several sensors to evaluate disparities in measurements. The results obtained from all the detectors are used to develop a rule-based isolation method that accurately identifies the source of the abnormal sensor. The effectiveness of the proposed architecture is further described by incorporating actual vehicle data, which also stress on helping protect autonomous vehicles from cyber risks.