Search
Close this search box.

Researcher Uses AI Camouflage to Outwit Flock Cameras

Baghdad: A 2009 Toyota Yaris, covered in a unique vinyl pattern, was recently driven past a Flock surveillance camera in a Las Vegas parking lot, challenging the camera’s object-detection capabilities. According to Iraqi News Agency, the unusual design was not merely decorative but part of a year-long project spearheaded by Bill Swearingen, founder of cybersecurity firm SIXCYBER. The initiative, known as NoRecognition, aimed to create adversarial designs capable of disrupting the function of object-detection software.

Despite the camera still recording video that is stored on a server, the classification layer-the crucial algorithm that identifies objects such as cars or license plates-fails due to Swearingen’s designs. These patterns exploit inherent weaknesses in neural networks, functioning as optical illusions specifically designed for machine processing rather than human perception. No physical alterations, such as spray paint, are used; instead, the designs rely on geometric configurations to disrupt machine learning models.

Swearingen asserts that these patterns successfully circumvent all 11 open-source detection algorithms he tested, impacting systems associated with Flock license plate readers, Axon body cameras, and Clearview AI. He emphasizes the importance of privacy as a fundamental right, positioning his project as a means for individuals to evade automated surveillance.

The NoRecognition project extends beyond vehicles, as it plans to offer printed clothing, including hoodies and T-shirts, as well as vehicle skins through crowdfunding. This development draws parallels to CV Dazzle, an earlier experiment by artist Adam Harvey, which used makeup to confuse facial recognition systems. However, Swearingen’s approach leverages millions of automated iterations rather than artistic intuition, marking a significant advancement in scale.

The ongoing battle between surveillance technology and privacy advocates is reminiscent of age-old security challenges. As surveillance vendors enhance their models, Swearingen plans to continuously refine his patterns, perpetuating the cycle of innovation and counteraction.