noRecognition Beats 11 Camera Algorithms, Foils Flock at Def Con
Bill Swearingen’s noRecognition patterns beat 11 camera algorithms and hid a Toyota Yaris from Flock detection in a live Def Con test on Friday.
Summary
TechCrunch reported on August 9, 2026, that Bill Swearingen’s noRecognition project can generate patterns on demand that disrupt automated identification of covered people, faces, objects and vehicles by commonly deployed surveillance cameras and license plate readers. The patterns do not prevent video recording, but suppress detection alerts. After roughly 31 million tests over one year, his reinforcement learning system, developed from a proof-of-concept lab and expanded with community-supplied computing power, defeated all 11 open-source algorithms tested, including software powering Flock license plate readers, Axon body-worn cameras and cameras running Clearview AI. It now generates a mathematically improved pattern every minute.
In the project’s first public real-world test on Friday at Def Con in Las Vegas, Donut Media helped cover a 2009 Toyota Yaris with a pattern. Swearingen said it defeated detection by a Flock camera, although the wheels remained challenging; Donut Media plans to release video within weeks. The demonstration provides early evidence that adversarial patterns can counter algorithmic surveillance outside a laboratory.
Swearingen, a Kansas City cybersecurity professional and SecKC co-founder, began the project after surveillance cameras placed only feet apart made him fear protest participants could be tracked while exercising constitutional rights. He also objected to government facial recognition using driver’s licenses without consent. A crowdsourcing campaign seeks to sell high-resolution T-shirts and hoodies that work at a distance, with vehicle skins potentially following. He is withholding the strongest designs so camera makers cannot quickly defeat them, while continued failures further train the model.
Positives
- 31 million tests over one year produced patterns capable of disrupting automated identification of covered people, faces, objects and vehicles.
- All 11 open-source algorithms tested were defeated, including software associated with Flock, Axon and Clearview AI camera systems.
- Friday’s Def Con demonstration showed a patterned 2009 Toyota Yaris evading Flock camera detection outside the laboratory.
- One new pattern is generated every minute, with each batch mathematically improving on its predecessor.
- Crowdsourcing could bring high-resolution patterns to T-shirts, hoodies and eventually vehicle skins.
Risks & concerns
- The patterns do not stop cameras from recording footage; they only interfere with automated identification and alerts.
- The 2009 Toyota Yaris demonstration exposed difficulty concealing the vehicle’s wheels from detection.
- One public Def Con test constitutes only early evidence that noRecognition will work reliably across public surveillance systems.
- Swearingen is withholding his strongest patterns because camera manufacturers could adapt their algorithms to defeat published designs.
- Cameras placed only feet apart made Swearingen fear protest participants could be tracked while exercising constitutional rights.
