A Public Test for Anti-Detection Patterns
A privacy-focused project called noRecognition is testing whether adversarial visual patterns can stop surveillance systems from automatically detecting people, faces or vehicles. TechCrunch reports that creator Bill Swearingen has run about 31 million tests and demonstrated a patterned vehicle at Def Con in Las Vegas. The claim is not that cameras stop recording, but that automated detection alerts may fail when the pattern covers the target.What noRecognition claims to do
noRecognition is presented as a way to interfere with algorithmic detection rather than with camera recording. According to TechCrunch, Swearingen says the computer-generated patterns can be applied to clothing or objects and prevent some commonly deployed surveillance cameras and license plate readers from detecting what the pattern covers.That distinction matters. A camera may still capture video, but the analytical layer that identifies a person, vehicle, plate or face may not flag the footage for attention. In Swearingen's framing, the covered subject becomes harder to pull out of large volumes of footage unless an operator already knows where to look. The practical implication is a direct challenge to automated surveillance workflows, not a cloak of physical invisibility.
A Def Con test moved the claim beyond a lab
TechCrunch reports that noRecognition had its first public test Friday at the Def Con cybersecurity conference in Las Vegas. The test used a 2009 Toyota Yaris covered with one of Swearingen's newer patterns to see whether it would be detected by a Flock camera.Swearingen told TechCrunch, "We proved it was effective," while also saying that the wheels were a challenge. The report says Donut Media helped with the test and that video of the demo is expected in the next few weeks. The public demonstration gives the project more news value than a private laboratory claim, though the available evidence remains limited to TechCrunch's report and Swearingen's account of the test.
How the pattern work is described
Swearingen told TechCrunch that the project evolved from a proof-of-concept lab into a reinforcement learning model that trains itself on which patterns work against specific camera algorithms. He described the process as teaching the model "how to paint," with failed patterns becoming feedback for later attempts.The system reportedly tested patterns against 11 open source detection algorithms, including software that TechCrunch says powers Flock license plate readers, Axon body-worn cameras and cameras running Clearview AI. Swearingen says the model now creates new patterns every minute, with each batch mathematically better than the last. For readers, the key point is not a reproducible method, but the broader security lesson: machine vision systems can be sensitive to specially designed visual inputs.
Why the privacy argument matters
Swearingen frames noRecognition as a privacy tool. He told TechCrunch, "Privacy is a fundamental right," and described the patterns as a way for people to "opt out of being tracked." He said his interest grew from seeing many surveillance cameras in his town and from concerns about people attending protests under pervasive camera coverage.The report places the project inside a longer line of attempts to counter facial recognition and object detection, including apparel and eyewear projects with mixed effectiveness. The policy implication is sharper than a fashion experiment: if surveillance networks increasingly depend on automated alerts, tools that degrade those alerts may force a debate over consent, public-space monitoring and the reliability of algorithmic identification.
Limits, incentives and likely countermeasures
The project is still early. TechCrunch reports that Swearingen is keeping his strongest patterns off the internet because he does not want camera makers to adapt against them, and that the work continues as his models generate new patterns. That suggests an adversarial cycle in which pattern designers and surveillance vendors each respond to the other's changes.There are also practical limits. The Def Con car test reportedly had trouble with wheels, and a single public demonstration does not establish performance across lighting conditions, camera angles, distances, weather, lens types or model updates. The most cautious reading is that noRecognition shows a credible pressure point in computer vision systems, while leaving open how durable and broadly effective the approach will be outside controlled or semi-controlled tests.
Conclusion
noRecognition is significant because it turns a technical weakness in machine vision into a public privacy argument. TechCrunch's report supports the claim that Swearingen has built and publicly tested adversarial patterns against surveillance detection, but it does not prove universal effectiveness against all cameras or conditions.For surveillance vendors, the project is a reminder that detection systems can face active countermeasures. For policymakers and the public, it raises a narrower but pressing question: if automated tracking becomes ordinary in public spaces, what lawful tools should people have to avoid being automatically identified?
Sources
Editorial Team - CoinBotLab
