Bill Swearingen has spent the last year engaged in a digital game of hide and seek, running more than 31 million tests to create a visual loophole in modern security systems. The result is a project called noRecognition, which produces computer generated adversarial patterns designed to blind the automated detection algorithms powering thousands of surveillance cameras and license plate readers across the United States. While these patterns do not stop a camera from recording video, they effectively scramble the artificial intelligence’s ability to recognize people, faces, or vehicles, preventing the software from triggering an alert.
The motivation behind the project stems from Swearingen’s concerns over the rapid expansion of algorithmic surveillance in cities like his own home of Kansas City. As a cybersecurity professional, he views the proliferation of cameras as an infringement on fundamental privacy rights, noting that citizens never opted into being tracked by government agencies or private firms. He recalls feeling hesitant to attend protests due to fear that facial recognition could be used to monitor those exercising their right to free expression, prompting him to build a tool that lets individuals opt out of this invisible net.
To achieve this, Swearingen developed a reinforcement learning model that essentially taught itself how to paint images that confuse machines. After iterating through countless failures against eleven different open source detection algorithms—including software used by major players like Clearview AI and Axon—the model can now generate highly effective patterns on demand. These designs target the mathematical vulnerabilities of the sensors, turning a recognizable object back into a needle in a haystack that requires manual human effort to find.
The theory moved into practice recently at the Def Con cybersecurity conference in Las Vegas, where Swearingen wrapped a 2009 Toyota Yaris in one of his latest patterns. The experiment proved successful, demonstrating that the vehicle could bypass detection by Flock license plate reader cameras in a real world environment. Looking ahead, Swearingen plans to bring these designs to the public through fashion items like t shirts and hoodies, aiming to blend aesthetic appeal with functional anonymity for anyone wishing to reclaim their privacy in public spaces.