New 'adversarial' patterns can evade surveillance cameras
Computer-generated patterns disable facial recognition systems.
1A new weapon against surveillance systems
Bill Swearingen has been developing computer-generated patterns for over a year. When applied to clothing or objects, these patterns successfully evade the automatic detection systems of license plate readers and surveillance cameras commonly used on American streets. After more than 31 million tests, Swearingen's noRecognition project can now produce these patterns on demand.
2How algorithmic surveillance works
Modern surveillance cameras can automatically identify activities in their recorded footage. They are used in a wide range of applications, from tracking speeding vehicles' license plates to identifying suspected criminals with facial recognition systems. However, these systems have variable success rates and can sometimes produce terrifying results. The detection algorithms powering these cameras sift through vast amounts of footage to pinpoint activities of interest.
3How the patterns work
Swearingen's computer-generated patterns do not block surveillance cameras from recording video. Instead, they disrupt the camera's ability to recognize objects, people, or faces, preventing any detection alerts from being triggered. By blocking the camera's ability to detect what the pattern covers, the person becomes a 'needle in a haystack' once again.
4Defending the right to privacy
Swearingen states, 'Privacy is a fundamental right,' describing these patterns as a way for people to 'opt out of being tracked.' As a cybersecurity professional living in Kansas City, he is acutely aware of the privacy and security risks posed by surveillance. He notes that his town is saturated with surveillance cameras, sometimes placed just a few feet apart.
5Successful real-world test
At the Def Con cybersecurity conference in Las Vegas, Swearingen conducted his first real-world test. He covered a 2009 Toyota Yaris with one of his newest patterns to see if it would evade detection by a Flock camera. The test was successful, though the wheels posed a challenge. A video of the demo will be released in the coming weeks.
6Project grows with community support
While developing his project, Swearingen benefited from hardware contributions by a broad community. What began as a proof-of-concept test lab eventually evolved into a reinforcement learning model. This self-contained system can teach itself which patterns work and which don't against specific camera algorithms. Through repeated trials, the model eventually defeated all 11 open-source detection algorithms it was tested against.
7Next steps
Following the successful demo in Las Vegas, Swearingen's next goal is to get these patterns into the hands of people who want them. The noRecognition project has launched a crowdfunding campaign to finance the sale of early merchandise featuring the patterns, from T-shirts to hoodies. Plans are also underway to offer pattern-printed skins for vehicles in the future. Swearingen aims to enable people to avoid algorithmic detection in public spaces.
8Resistance against surveillance technology
Swearingen's work represents a new phase in resistance against surveillance technologies. Previously, some art projects and clothing brands have offered garments designed to evade facial recognition systems. However, Swearingen's patterns stand out as one of the most effective solutions in this field. The project aims to help individuals feel secure while exercising their constitutional rights.