Seeing is Deceiving: Systematic Vulnerability Analysis of LiDAR-Based Autonomous...
summary
Systematic tests show planar mirrors cause two reproducible failures: Object Addition (phantom obstacles) and Object Removal (real obstacles vanish). Experiments use an Oyster OS1 128‑beam LiDAR on an Autoware stack; a 0.18 m² mirror yields 74% “Vehicle” classification and forces emergency braking, and injected CARLA artifacts stop an ego vehicle from 25 km/h in 1 s causing a rear‑end crash. Physics‑informed models fit with R²≈0.85–0.98 and a real‑time CARLA-based simulation injector is implemented; thermal and intensity checks show clear limitations.
AI-written summary · about
chapters (12)
- 0:00Autonomous vehicle perception
- 1:15LiDAR vulnerabilities and attacks
- 2:19Mirror LiDAR artifacts
- 3:46Mirror-based attack analysis
- 4:27Object removal demo
- 5:08Object addition setup
- 6:56Sensor-agnostic mirror flaws
- 7:17End-to-end system impact
- 9:46Artifact modeling
- 10:31CARLA simulation
- 11:35Defense strategies
- 13:43Future challenges
related talks
01h 06mSelf-Driving and Connected Cars: Fooling Sensors and Tracking Drivers
40mSecuring Tesla and Mobileye from Split-Second Phantom Attacks
13mBeyond the Glow: Understanding Luminescent Marker Behavior Against Autonomous...
11mWIP: Evaluating the End-to-End Impact of False Localization Attacks on vSLAM-...
10mWIP: Evaluation of Threats and Impacts of HD Map Tampering Attacks in Autonomous...
09mWIP: Hijacking Attacks on UAV Follow-Me Systems in Realistic Scenarios
11mWIP: Understanding the Mechanisms Behind NDT-Based Localization Vulnerabilities...
06mLightning Talk: PEPPAR on DS: Platform for Evaluating Physical-Layer and...