Paper accepted at ICRA 2025: “SLAMSpoof”
Full self-driving services rely on map information to recognize lane shapes and the positions of traffic lights and signs, which demands centimeter-level localization accuracy. Today only LiDAR can deliver that accuracy, yet LiDAR is known to be vulnerable to spoofing attacks that overwrite its measurements with malicious laser emissions.
This work presents SLAMSpoof, the first practical LiDAR spoofing attack on localization systems for self-driving. It identifies effective attack positions using the scan matching vulnerability score (SMVS), a point-wise metric of susceptibility to spoofing. In real-world experiments on ground vehicles, the attack induced position errors greater than 4.2 m — more than a typical lane width — against all three popular LiDAR-based localization algorithms. The paper also discusses potential countermeasures.
[Paper]
Rokuto Nagata, Kenji Koide, Yuki Hayakawa, Ryo Suzuki, Kazuma Ikeda, Ozora Sako, Qi Alfred Chen, Takami Sato, Kentaro Yoshioka, “SLAMSpoof: Practical LiDAR Spoofing Attacks on Localization Systems Guided by Scan Matching Vulnerability Analysis,” IEEE International Conference on Robotics and Automation (ICRA 2025), Atlanta, GA, USA, May 2025.