Bill Swearingen Demonstrates Pattern That Bypasses Surveillance Cameras

Cybersecurity researcher Bill Swearingen demonstrated a computer-generated pattern at the Def Con conference in Las Vegas that successfully prevents automated license plate readers and surveillance cameras from detecting vehicles and objects.

The system is the result of roughly 31 million tests run over the past year by Bill Swearingen, a co-founder of the SecKC cybersecurity meetup based in Kansas City. Swearingen set out to build a tool that could allow people to opt-out of being tracked as municipal surveillance expands across the United States. Rather than jamming video signals or blocking cameras from recording footage, the computer-generated designs scramble the detection algorithms that power modern surveillance equipment, preventing the software from triggering alerts on people, faces, or vehicles.

Teaching a Reinforcement Learning Model How to Paint

Swearingen’s proof-of-concept began as a manual test lab evaluating open-source video camera detection algorithms one by one. Over time, the project evolved into a self-contained reinforcement learning model capable of training itself on which patterns succeed and which fail.

In a call with reporters, Swearingen explained that he essentially taught the model how to paint, letting the software iterate continuously until it defeated multiple algorithms simultaneously. The model eventually developed recipes capable of bypassing 11 open-source detection algorithms, including the software running Flock license plate readers, Axon body-worn cameras, and systems operated by Clearview AI. The model now generates fresh, mathematically refined patterns every minute.

Real-World Testing at Def Con in Las Vegas

The live test proved that computer-generated adversarial patterns can defeat automated detection in real-world conditions rather than just inside a lab simulation.

Swearingen noted that his motivation stemmed from feeling uncomfortable attending a public protest last year amid pervasive municipal cameras. He observed that while he personally had not faced discrimination as a middle-aged white man in the central United States, many individuals exercising their constitutional rights to free expression might not feel safe or comfortable doing so under constant algorithmic watch.

Eyewear Designers and Retail Brands Push Back

While computer scientists target infrastructure and license plate readers, a parallel movement is growing among fashion designers and eyewear brands producing adversarial styles to scramble surveillance cameras. These wearable anti-surveillance products arrive as tech companies introduce smart glasses equipped with real-time facial recognition and recording capabilities.

Nate Troxell, the company’s general manager, explained that the lenses are engineered to guard against infrared and blue light rays in addition to UV radiation. We’re making a firewall for your face, Troxell told reporters, adding that the underlying technology originated from cannabis greenhouses designed to block artificial light exposure.

High Fashion Meets Biometric Defiance

Gramercy-based designer Kerin Rose Gold, owner of the luxe eyewear brand A-Morir, has also entered the anti-surveillance market. Gold, whose custom pieces have outfitted performers like Lady Gaga and Rihanna, debuted a bespoke 3D-printed line named Handle With Care priced between $425 and $725. The designs draw inspiration from the United States Navy’s historic dazzle camouflage from the early 1900s, using stark black-and-white motifs to disrupt facial recognition sensors.

Bill Swearingen Demonstrates Pattern That Bypasses Surveillance Cameras — Photo: Techcrunch

Gold also offers a $625 design modeled after Japanese anime eyes, created initially for a Vogue photoshoot. She noted that a model trying to unlock her phone during a break found herself unable to do so while wearing the frames, prompting the designer to recognize their effectiveness against biometric technology. Big-box retailers have similarly joined the trend, with Zenni Optical offering an ID Guard model designed to stop data peepers.

These commercial products build upon a broader category of adversarial clothing designed to exploit vulnerabilities in computer vision systems. Previous efforts include the musician M.I.A.’s Ohmni anti-tracking line, featuring a $65 tinfoil hat meant to block electromagnetic waves, and Italian knitwear designer Rachele Didero’s 2019 launch of Cap_able, which uses patterned textiles to thwart facial recognition software.

Keep reading