A cybersecurity researcher has publicly demonstrated a technique using computer-generated patterns to prevent surveillance cameras from detecting vehicles, including those equipped with AI-powered license plate readers. Bill Swearingen, founder of SIXCYBER, showcased his "noRecognition" system at the recent DEF CON security conference in Las Vegas.
Swearingen's noRecognition is a reinforcement learning model that generates specific visual patterns. These patterns are designed to defeat the object detection algorithms used by surveillance cameras. While the cameras still record footage, and a human observer would clearly see a vehicle, the software designed to identify objects, log license plates, or trigger alerts fails to register the car.
The researcher spent a year developing the system, conducting 31 million tests. His method involved setting up a feedback loop where generated patterns were tested against camera detection software. If a pattern was detected, the system adjusted and iterated until a pattern was found that evaded detection.
During the DEF CON demonstration, Swearingen wrapped a 2009 Toyota Yaris in one of his newly generated patterns. He then drove the vehicle past a live Flock camera. The camera successfully recorded the car, but its detection software did not log any vehicle information, confirming the technique's effectiveness. Swearingen noted that the vehicle's wheels presented a particular challenge for the system.
The noRecognition system has been tested against 11 open-source detection algorithms. Swearingen claims it is effective against all of them, including the software used by Flock license plate readers, Axon body-worn cameras, and cameras employing Clearview AI's facial recognition system. He continues to generate new patterns constantly, keeping the most potent ones offline to prevent camera manufacturers from training their systems against them.
Swearingen initiated the project due to privacy concerns, specifically discomfort with being tracked while attending a protest. He views the patterns as a means for individuals to "opt out of being tracked," emphasizing privacy as a fundamental right.
The patterns are applied to the body of the car, deliberately avoiding the license plate itself, as obscuring a number plate is illegal. The legality of covering a car's bodywork with such a pattern, designed to circumvent surveillance software, remains an open question.
Flock Safety sells automated license plate recognition cameras to police departments across the United States. These cameras photograph passing vehicles and cross-reference their plates in real-time against law enforcement databases. The company's network has expanded significantly, and it has reportedly proposed integrating 350,000 Uber and Lyft dashcams into its nationwide scanning efforts. This expansion has drawn criticism from privacy advocates. Concerns also exist regarding the accuracy of Flock cameras, with reports of innocent drivers being subjected to stops due to incorrect plate matches.
