A researcher announced the creation of patterns designed to prevent detection by cameras
Cybersecurity expert Bill Swiringen has developed the noRecognition project, which generates computer-generated patterns for clothing, objects, and cars. According to him, these patterns can prevent video surveillance systems from automatically detecting people, faces, objects, or vehicles. TechCrunch reports on this.
Swiringen claims to have conducted approximately 31 million tests over the course of a year. The patterns do not prevent the camera from recording video, but, as intended by the developer, they complicate the work of algorithms that analyze images and send automatic alerts about detected objects, people, or faces.
The researcher began by testing against open-source object detection algorithms and later created a reinforcement learning model. The system repeatedly tested the generated patterns and modified them whenever the algorithm still managed to detect an object. According to Swiringen, the model found variants capable of countering all 11 algorithms he tested, including the software used in Flock license plate readers, Axon body cameras, and cameras powered by Clearview AI.
For more breaking news, follow the UA.News Telegram channel.
The first public test took place at the Def Con cybersecurity conference in Las Vegas. One of the new patterns was applied to a 2009 Toyota Yaris and tested in front of a Flock camera. Swiringen stated that the test confirmed the approach’s effectiveness, although the car’s wheels proved to be a challenging element to camouflage. Donut Media assisted with the test and plans to release a demonstration video in the coming weeks.
The project is also running a crowdfunding campaign to produce T-shirts and hoodies featuring these patterns. Swiringen said he is not publishing the most effective designs online so that camera manufacturers cannot develop countermeasures against them.
