Researcher uses machine learning to hide vehicles from Flock cameras
21 days ago
- Cybersecurity expert Bill Swearingen developed noRecognition, a reinforcement learning model that generates patterns to defeat AI-powered surveillance cameras like Flock license plate readers.
- The patterns do not block human viewing but fool detection software that logs license plates and triggers alerts.
- Swearingen demonstrated the technique at DEF CON by driving a pattern-wrapped car past a Flock camera, which recorded the vehicle but failed to log it.
- The system has been tested against 11 open-source detection algorithms, including those used by Flock, Axon body cameras, and Clearview AI, and beats all of them.
- Swearingen continues generating new patterns every minute and keeps the strongest ones offline to prevent camera manufacturers from training against them.
- Flock cameras are controversial for their rapid expansion, inaccuracies that lead to innocent drivers being pulled over, and privacy concerns.
- The patterns cover the car body but not the license plate to avoid illegality, as obscuring a plate is illegal.