AI Cameras Foiled by Fashion?

A new startup, noRecognition, is challenging pervasive AI surveillance with an innovative approach to car customization. Founded by cybersecurity researcher Bill Swearingen, the company has developed computer-generated patterns that can render objects nearly invisible to AI-powered detection systems. These designs are intended to be applied as car wraps, specifically to confound Flock's automatic license plate readers (ALPRs) and other surveillance technologies that collect vast amounts of personal data.

Swearingen demonstrated a proof-of-concept vehicle at the Def Con cybersecurity conference in Las Vegas, showcasing a car adorned with a distinctive blue and yellow abstract wrap. The company reports that its patterns have undergone stress-testing against eleven open-source detection algorithms, including software utilized by Flock cameras and Axon body cameras. This initiative stems from a growing concern over the proliferation of surveillance cameras and a desire to protect fundamental privacy rights.

The underlying strategy employs adversarial media, a technique that injects malicious data or, in this case, specific patterns of shapes and colors, into images or videos. These patterns exploit algorithmic vulnerabilities in AI, causing misclassification or rendering objects undetectable to computer vision systems. Similar methods, known as 'cv dazzle,' have been used by protesters who apply surreal face paint designs to evade facial recognition software.

While this anti-AI customization shows promise, its effectiveness is not guaranteed across all scenarios. Each pattern requires specific testing against distinct detection algorithms, and AI technology is continuously advancing. The success of noRecognition's approach will depend on its ability to adapt and maintain efficacy against evolving surveillance capabilities.