Researchadversarial examplescomputer visionautonomous vehiclestesla
Keen Lab Reveals Tesla Autopilot Vulnerabilities In Tests
9.1
Relevance Score
Tencent Keen Security Lab in demonstrations and a technical report details three experimental vulnerabilities in Tesla's Autopilot, showing that adversarial images can trigger auto-wipers, roadside stickers can induce lane misclassification, and a compromised Autopilot can allow steering control via a wireless gamepad. Tesla said some issues were fixed in 2017–2018 updates and argued the demonstrations used unrealistic conditions, while Keen Lab published videos and a paper.



