Paper accepted at IROS 2026: “MirrorDrift”
Prior physical attacks on LiDAR SLAM rely on LiDAR spoofing via external signal injection, which requires sensor-specific timing knowledge and is increasingly blocked by modern defenses such as timing obfuscation and injection rejection.
This work turns to specular reflection as an injection-free alternative and presents MirrorDrift, which uses an actuated planar mirror to create ghost points and systematically bias scan-matching correspondences. By optimizing mirror placement, alignment, and actuation, it increased average pose error by 6.1x over random placement in simulation, degrading three SLAM systems to 2.29-3.31 m mean APE, and induced localization errors of up to 6.03 m in real-world experiments on a modern LiDAR with state-of-the-art interference mitigation. The work will be presented at IROS 2026 in Pittsburgh in September 2026.
[Paper]
Rokuto Nagata, Kenji Koide, Kazuma Ikeda, Ozora Sako, Shota Horie, Kentaro Yoshioka, “MirrorDrift: Actuated Mirror-Based Attacks on LiDAR SLAM,” IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026), Pittsburgh, PA, USA, September 2026.