Enforcing Safety in Cooperative Perception of Autonomous Driving Systems through Logistic Chaos Map-based End-to-End Encryption

Abstract

Collaboration among multiple cyber-physical systems (CPSs) requires improved safety, reliability, and performance. Collaborative CPSs share common goals and collaborate to achieve them. Connected and autonomous vehicles (CAVs) are typical examples of collaborative CPSs. The Cooperative perception in CAVs is an emerging technology that enables the CAVs to share their local perception with others, thereby improving efficiency and road safety. However, in cooperative perception, malicious vehicles may send phantom vehicle information, and additionally, vehicles may unintentionally be malicious due to faulty sensors. These issues pose serious driving hazards as they can incur traffic accidents. Therefore, this article uses logistic chaos map-based end-to-end encryption techniques to avoid malicious vehicle information in cooperative perception. The cooperative perception is achieved via sharing the camera sensor image frames between two vehicles. Using the CARLA simulator, we demonstrated the real-time logistic chaos map-based encryption in the cooperative perception of CAVs. Unlike existing baseline approaches such as Cooper and F -Cooper, in our cooperative perception, we first encrypt the image frames before sharing and then decrypt the image frame at receiving end to avoid malicious information. The experimental results, such as the histogram, adjacent pixel correlation, and key sensitivity analysis, demonstrated that cooperative perception using logistic map-based encryption is safer and more secure than existing methods. In addition, our cooperative perception system increased the detection rate up to two times than the individual perception system of CAVs.