A 3D Point Attacker for LiDAR-Based Localization

Shiquan Yi, Jiakai Gao, Yang Lyu, Lin Hua, Xinkai Liang, Quan Pan

Abstract: The safety and security issues of autonomous navigation function become the main obstacles that hinder the widespread applications of self-driving cars and unmanned systems. In this paper, we investigate the vulnerability of LiDAR-based localization methods to adversarial attacks. Specifically, we developed a feature-based spoofing attack strategy to degrade the localization performance of LiDAR-based localization algorithms. Reflecting on the vulnerability, we additionally provide a resilient strategy to defend existing LiDAR-based localization methods against this attack. The proposed attack strategy is tested on the KITTI dataset to illustrate its effectiveness.

External IDs:dblp:conf/icca/YiGLHL024