NoRecognition: AI Camouflage That Foils Surveillance Cameras, Shown With Flock
A Kansas City researcher created noRecognition patterns that use adversarial machine learning to confuse AI-based detectors in surveillance systems like Flock. The patterns don’t block cameras or blur images; they prevent the detection layer from recognizing vehicles or plates, so footage is recorded but not logged as a car. Demonstrated at Def Con with a wrapped 2009 Toyota Yaris passing a Flock camera, the patterns are generated by reinforcement learning, update continuously, and are kept offline to avoid easy countermeasures. The work frames privacy as a right to opt out of automated tracking, though legal questions remain about driving publicly with such camouflage.
