Inside noRecognition: The Anti-Surveillance Project That Used 31 Million AI Tests to Outsmart Flock, Axon, and Clearview Cameras

The automated surveillance systems that scan license plates, identify faces, and track vehicles have long seemed like an unstoppable fact of modern life. But a Kansas City cybersecurity researcher has built something that claims to stop them cold—without blocking a single camera.

Bill Swearingen's noRecognition project uses reinforcement learning to generate adversarial visual patterns that, when printed on fabric or vinyl, scramble the object-detection and classification software running on widely deployed surveillance cameras. After roughly 31 million automated tests spanning a year, the system converged on designs that defeated 11 open-source detection algorithms in the lab. Then, on Friday, August 7, 2026, Swearingen publicly proved it worked in the real world: a 2009 Toyota Yaris covered in one of his patterns drove past a Flock license plate reader at the Def Con cybersecurity conference in Las Vegas and escaped automated detection entirely .

Here is how noRecognition works, what the Def Con demo actually showed, and how anyone can now buy these patterns as clothing.

What is noRecognition?

noRecognition is an anti-surveillance research project by Bill Swearingen, a Kansas City cybersecurity professional and founder of the SecKC meetup and SIXCYBER. It produces adversarial visual patterns designed to prevent automated surveillance systems from detecting people, faces, vehicles, and objects . The patterns are printed onto fabric or vinyl and applied to clothing or vehicle surfaces. They do not block ordinary video recording — they specifically fool the object-detection and classification software running on those cameras .

Swearingen described his approach to TechCrunch: the patterns exploit the classification layer of neural networks, creating a visual signal that the model cannot parse into a recognizable category . To a human observer, the pattern looks like an abstract, chaotic design. To an AI, it is effectively invisible.

How the reinforcement learning system trained over 31 million tests

Swearingen built a reinforcement learning model that iteratively generated patterns and tested them against 11 different open-source computer vision detection algorithms . With each test, the model learned which visual features caused detection to fail and refined the pattern accordingly.

After approximately 31 million tests (some sources report 31.7 million) run over roughly a year, the model converged on designs that consistently caused those algorithms to fail to classify whatever the pattern covered . Swearingen has described the process as teaching the model "how to paint" — the system learned to generate textures that maximize confusion in the detection software .

In controlled digital tests using real extracted model weights, a full-coverage noRecognition texture drove the f-YOLOv5 surveillance detector to a 61.7% non-detection rate across 240 held-out garment images at the 0.25 detection threshold, with an occlusion-subtracted lift of +0.537 over the untextured baseline .

What the patterns defeated in lab testing:

• Flock license plate readers
• Axon body cameras
• Clearview AI facial recognition systems
• All 11 open-source detection algorithms in Swearingen's test suite

The Def Con 34 demo: a 2009 Toyota Yaris that cameras couldn't see

On Friday, August 7, 2026, Swearingen conducted the first public real-world test of a noRecognition pattern at the Def Con cybersecurity conference in Las Vegas . Working with automotive media outlet Donut Media, he covered a 2009 Toyota Yaris in one of his computer-generated patterns and drove it past a Flock license plate reader .

Swearingen reported that the Yaris escaped automated detection entirely — the Flock camera could not identify the vehicle or read its license plate, while normal video recording continued without issue . The wheels reportedly presented a challenge, but the overall test was deemed effective .

Important caveat about the demo: Independent verification of the claim has been limited. TechCrunch noted that the result is narrower than a vehicle becoming "invisible" — the camera still recorded footage, while the demonstration video and logs are not public . The physical-world record currently stands at one Def Con test against a single camera model, and proof remains a point of discussion .

How to buy noRecognition patterns: T-shirts, hoodies, and neck buffs

Swearingen's stated goal is to turn noRecognition into a practical tool for public privacy — a way for people to opt out of algorithmic surveillance without needing technical expertise .

He launched a Kickstarter campaign that went live simultaneously with his research presentations at Black Hat and Def Con in early August 2026 . He announced on LinkedIn: "Tees, hoodies, and neck buffs printed with the patterns that actually work" — released at the same moment as the research .

Current and planned merchandise:

Product	Status
T-shirts	Available via Kickstarter campaign
Hoodies	Available via Kickstarter campaign
Neck buffs / gaiters	Available via Kickstarter campaign
Vehicle skins / wraps	Planned, following the clothing line

As of mid-August 2026, the Kickstarter campaign had raised more than $40,000 . Backers who pay a premium can request custom patterns that were "generated, tested, and never released, shown, or printed for anyone else" . The vehicle skin application is described as a next step once the clothing line is underway .

Important legal and practical caveats

The project remains a privacy research initiative rather than a proven commercial product . The lab testing showed success against specific algorithms, but real-world performance against the full range of deployed surveillance systems has not been independently audited or guaranteed .

Swearingen has also acknowledged legal gray areas: driving a wrapped car raises state-by-state questions about license plate obstruction, and his patterns explicitly avoid covering license plate areas .

For now, noRecognition offers a glimpse of a possible future where individuals can opt out of algorithmic surveillance by wearing the right pattern. But Swearingen himself, along with multiple news outlets, has been careful to note that this is research — not a magic cloak — and that the evidence base remains a single, non-publicly-documented camera test .