Bill Swearingen AI surveillance evasion research

Researcher Bill Swearingen has developed ‘noRecognition’, an experimental project designed to disrupt automated surveillance and machine-learning algorithms. The technology uses specialized geometric and contrast patterns on clothing, objects, or vehicles to confuse computer vision software, causing it to classify targets as ‘background noise or empty space’ rather than making them invisible to human eyes.

Following extensive testing involving 31 million reinforcement learning simulations against 11 different systems, Swearingen demonstrated the technology at DEF CON 2026. In a demonstration conducted in collaboration with the YouTube channel Donut Media, a 2009 Toyota Yaris wrapped in the specialized patterns was driven past a Flock Safety automated license plate reader. The test successfully demonstrated that while the camera recorded the footage, the software failed to identify the object as a vehicle.

Testing confirmed the effectiveness of this adversarial machine learning approach against various systems, including Axon body cameras and Clearview AI facial recognition software. The patterns are designed to disrupt the object-detection layer of computer vision systems by providing enough geometric and contrast-based noise to prevent AI from recognizing and logging subjects such as people, faces, or vehicles.