Data-driven anomaly identification for cameras and lidars mounted on vehicles

Abstract

Cyber-physical systems (CPS) have advanced rapidly and fueled the growth of automated driving in electric vehicles (EVs) that rely on AI and machine learning for functions like image recognition and navigation. However, the integration of sensors such as cameras, LiDARs, and radars makes these systems vulnerable to cyber attacks, posing potentially fatal threats. This project focuses on data-driven approaches to detect anomalies and mitigate these attack vectors in autonomous vehicles. By analysing data from cameras and LiDARs, and employing machine learning techniques, the aim is to promote early detection and lower the risk of threats like LiDAR spoofing and jamming attacks can bring. As the industry pushes toward fully autonomous driving, robust cybersecurity measures are essential to ensure the safety and reliability of these systems.