Testing the 'Invisible' Pattern: Can Adversarial Art Actually Blind Surveillance?
AI-generated image · Bay Street Wire
Bill Swearingen's noRecognition project claims to defeat the algorithms powering Flock and Clearview AI, moving from theoretical labs to a real-world Def Con demo.
I've always been skeptical of 'privacy fashion'—those high-concept clothing lines that claim to hide you from the grid but usually just make you look like a glitchy piece of modern art. But the latest developments from Bill Swearingen's **noRecognition** project suggest we might be moving past theoretical curiosity into something that actually works on the street.
As TechCrunch first reported, Swearingen—a cybersecurity professional and co-founder of the SecKC meet-up in Kansas City—has spent the last year attempting to create computer-generated patterns that can block the detection capabilities of common surveillance cameras and license plate readers. After roughly 31 million tests, Swearingen says he can now produce patterns on-demand that prevent these systems from detecting the people or vehicles they cover.
It is important to clarify the tech here: the patterns do not stop a camera from recording video. Instead, they scramble the detection algorithms that trigger alerts for specific objects, faces, or license plates. In Swearingen's words to TechCrunch, the goal is to make the subject a "needle in a haystack" again, allowing individuals to opt-out of being tracked by algorithmic surveillance.
Swearingen's process involved building a reinforcement learning model that essentially taught itself "how to paint." By iteratively testing patterns against 11 different open-source detection algorithms, the model refined its output until it could defeat multiple systems simultaneously. TechCrunch reports that these patterns are effective against software used by **Clearview AI**, **Axon** body-worn cameras, and **Flock** license plate readers.
To see if this held up outside the lab, Swearingen conducted a public test at the Def Con cybersecurity conference in Las Vegas. Partnering with **Donut Media**, he wrapped a 2009 Toyota Yaris in one of these adversarial patterns to see if it could evade a Flock camera. Swearingen told TechCrunch the test proved effective, though he noted that the wheels presented a specific challenge.
As for the rollout, Swearingen is launching a crowdsourcing campaign to fund merchandise, including hoodies and T-shirts, with plans for vehicle skins in the future. He aims for the patterns to be high-resolution enough to work from a distance while remaining aesthetically fashionable. (Opinion: If these actually hold up against the heavy hitters like Clearview, we're looking at a genuine shift in the hardware arms race between privacy advocates and surveillance tech.)