New patterns can make objects invisible to surveillance cameras

SecKC Founder Bill Swearingen has developed computer-generated patterns that can block surveillance cameras from detecting objects, people, or vehicles. After approximately 31 million tests, Swearingen's project, called noRecognition, can now produce patterns on demand that prevent commonly deployed license plate readers and surveillance cameras from identifying what the pattern covers. This technology aims to allow individuals to opt out of automatic detection and algorithmic surveillance, with further coverage provided by TechCrunch.

Swearingen's patterns do not prevent cameras from recording but instead scramble their ability to identify specific objects, effectively making them invisible to detection algorithms. This research builds upon previous efforts in adversarial art and clothing designed to confuse facial recognition systems. Swearingen utilized a reinforcement learning model, which he described as teaching the model "how to paint," to iteratively refine patterns that could defeat multiple open-source detection algorithms, including those used in Flock license plate readers, Axon body cameras, and Clearview AI.

A public demonstration at the DEF CON cybersecurity conference successfully showed a vehicle covered in a pattern being undetectable by a Flock camera. The project aims to make these patterns accessible to the public through merchandise and potential vehicle skins, with Swearingen continuing to develop stronger patterns by analyzing failures.

Source: TechCrunch