A cybersecurity researcher has developed a system, dubbed "Project noRecognition," that generates patterns designed to prevent surveillance cameras and automated license plate readers from recognizing individuals or vehicles. The project, led by Kansas City-based researcher Bill Swearingen, aims to disrupt the detection layer of surveillance software rather than physically obscuring objects from the camera lens.
Swearingen's research involved testing approximately 31 million patterns over the past year. His method uses a reinforcement learning system that learns from detection failures, adjusting patterns to improve their ability to evade multiple detection algorithms simultaneously. The core idea is not to hide objects from the camera itself, but to make the artificial intelligence software analyzing the footage unable to identify specific features like faces or license plates.
The project's dashboard, available at sandbox.norecognition.org, details the research, acknowledging that the goal of a single pattern defeating every detector is only partially met. The strongest validated result against a detector extracted from a real deployed surveillance camera achieved a 61.7% non-detection rate in digital simulations using held-out test subjects. However, the researchers emphasize the distinction between simulated results and real-world application, noting that most headline figures are explicitly labeled as digital.
A significant real-world test occurred at the DEF CON cybersecurity conference in Las Vegas. Swearingen, with assistance from Donut Media, covered a 2009 Toyota Yaris with one of his newest patterns and tested it against a Flock Safety camera, a type widely used for automated license plate reading in the United States. Swearingen stated that the test proved effective, though the vehicle's curved wheels presented a challenge, suggesting flat printed patterns are less effective on such surfaces. Donut Media is expected to release video of this demonstration in the coming weeks.
Swearingen initiated the project due to personal concerns about increasing surveillance in his hometown and the potential for tracking during protests. He is not publicly releasing his most effective patterns to prevent camera manufacturers from easily developing countermeasures. Instead, he plans to use crowdfunding to develop and sell printed products, such as T-shirts and hoodies, with vehicle wraps potentially becoming available later.
The practical utility of Project noRecognition as an everyday privacy tool remains to be fully determined. Its effectiveness will depend on how well the patterns perform on actual clothing and vehicles under various real-world conditions, including different weather and camera setups.
