An Efficient Method to Detect both Insertion and Removal Attacks in 3D LiDAR Point Clouds for Autonomous Vehicle Security

The development of autonomous vehicles (AVs) in recent years and the prospects of future advancements require the creation of solutions to protect the data used by these vehicles. Indeed, the operation of AVs relies on data from numerous sensors. The reliability and security of sensor data are major challenges in ensuring the successful development of AV technologies. LiDAR (light detection and ranging) is one of the key sensors that AVs use to understand their environment. This paper proposes a method to protect the data acquired by LiDAR in AVs, leveraging the physical coherence of the 3D LiDAR point clouds to detect both point insertion and removal attacks. Experimental results demonstrate the effectiveness of our method and its ability to achieve satisfactory detection accuracy for both insertion and removal attacks while maintaining a low execution time.