From adversarial patterns to infrared lights embedded in fabrics, a market for wearable products is challenging the biometric identification systems deployed across public spaces.
Cameras connected to facial recognition software are now present in airports, shops and streets across much of Europe. In the United Kingdom, as reported by the Guardian, a new generation of designers is responding with garments featuring so-called “adversarial patterns” – sequences of shapes, colours and repeated motifs designed to exploit the weaknesses of computer vision systems. The stated aim is to preserve anonymity in public spaces where automated surveillance has become the norm.
The British context makes the phenomenon particularly relevant: the Metropolitan Police in London operates real-time facial recognition and attributes roughly 2,500 arrests to the technology since the start of 2024. The country has 7.5 million government and private cameras in public spaces, excluding home security and dashcam devices. Critics, cited by Reuters, argue the system treats every passer-by as a potential suspect, undermining the presumption of innocence.
Among the companies active in this segment is Italian firm Cap_able Design, which produces T-shirts with adversarial patterns woven into the fabric: according to the company, object-recognition software identifies the garment as a dog or a giraffe rather than registering the face of the wearer. German firm Urban Privacy sells the Urban Ghost coat, fitted with built-in infrared lights invisible to the human eye but capable of saturating camera sensors. It also offers sweatshirts and tops with graphic motifs intended to confuse biometric analysis systems. US company Adversarial Apparel markets a clothing line described by the company itself as designed to “disrupt biometric profiling and refuse continuous data extraction”.
Those who prefer less conspicuous options can turn to anti-surveillance glasses made by Reflectacles (USA) and Sunphey Optical (China): both companies sell frames with reflective coatings and infrared filters that obscure the eyes, which most facial recognition algorithms treat as their primary reference point. Low-technology techniques – hats, face coverings, ordinary glasses – remain effective as well, since the software can only identify what the camera is actually able to capture.
The market must contend with rapid advances on the other side. A researcher who tested one garment for the Mozilla Foundation confirmed the product’s effectiveness but noted that continuously updated algorithms can render adversarial patterns obsolete quickly. One entrepreneur in the sector acknowledged that “algorithms are becoming increasingly difficult to fool.” The useful life of an anti-surveillance garment may therefore shrink to a few months before the targeted systems are updated.
The Metropolitan Police has announced no plans to reduce the use of real-time facial recognition on London’s streets.
