The evolving aesthetics of privacy

JOEL WILLIAMS joel.williams@photonics.com

Cameras are everywhere. They are in our pockets, on our computers, on traffic lights, and on the sides of and inside buildings. Surveillance, for better or worse, is becoming a part of our everyday lives and raises myriad philosophical, ethical, and personal privacy questions. In response to these increasing levels of surveillance, digital rights activists and artists have come up with unique methods of protest, namely through fashion and accessories.

Activists and hackers have been working on methods to fool facial recognition software for almost a decade, using stick-on jewelry and even sunglasses equipped with reflecting frames to trick facial recognition systems. In 2015, through a project called URME Surveillance, artist Leo Selvaggio created a prosthetic mask of his face. It was able to fool facial recognition software when others wore it.

Some methods can be a bit ostentatious (stick-on facial jewelry), a bit expensive (IR-blocking sunglasses), and maybe even a bit too police-attracting (Selvaggio’s mask).

Hacker and fashion designer Kate Rose uses a different method to protest surveillance. Her recently introduced line, Adversarial Fashion, features simple license plate prints on T-shirts, knapsacks, dresses, skirts, and crop tops. The line is designed to trigger automated license plate readers and load junk data into the monitoring systems, which have become ubiquitous in many cities and towns around the world.

Another clothing line designed to confuse recognition systems is HyperFace, which was developed by artist Adam Harvey in 2017. HyperFace textiles use patterns depicting many various-size face-like objects to lower facial recognition confidence scores by redirecting attention to the false faces. The line is an extension of Harvey’s earlier work on a 2011 project called CV Dazzle — a collaboration with makeup artists to use avant-garde hairstyling and makeup designs to negate facial recognition.

Adam Harvey’s HyperFace design uses false images of faces to distract facial recognition software and lower its confidence scores. Possible detections of faces (top), and the heat map for facial detections (bottom).

Rather than creating distractions or decoys, CV Dazzle used makeup to alter the expected light and dark areas of a face. With the expected recognition points broken up or obscured, neural networks are unable to compare and match the image. The project’s title comes from the term “dazzle camouflage,” a technique created by artist Norman Wilkinson during World War I to cover warships with geometric patterns designed to prevent enemy combatants from gauging a vessel’s speed, range, or heading.

Much of Harvey’s anti-surveillance work targets OpenCV, a popular open-source computer vision and machine learning software library.

Kate Rose’s Adversarial Fashion line features license plate patterns designed to load junk data into automatic license plate detector databases.

Another artist, Simone C. Niquille, had a project called REALFACE Glamouflage in 2013, which emblazoned garments with full and partial images of famous faces, such as Barack Obama and Britney Spears. The project targeted Facebook’s facial recognition software and was designed to lower the confidence score of Facebook’s system.

Though none of these methods are capable of taking out a security camera system outright, they are harmless methods of protest.

As Sir Francis Bacon said: “Fashion is only the attempt to realize art in living forms and social intercourse.”