Research papers · 2020.08
Towards Robust LiDAR-based Perception in Autonomous Driving: General Black-box Adversarial Sensor Attack and Countermeasures
Towards Robust LiDAR-based Perception in Autonomous Driving: General Black-box Adversarial Sensor Attack and Countermeasures
Overview
The study analyzes LiDAR spoofing vulnerabilities caused by neglected occlusion patterns. CARLO checks physical occlusion invariants, while SVF incorporates those features into learning; experiments evaluate defenses across targeted perception models.
Author-supplied source information
Author-supplied information on arXiv.
18 pages, 27 figures, to be published in USENIX Security 2020
Version: arXiv v1 — manuscript revision