HyperFace OpenCV Prototype©Adam Harvey
Fighting the surveillance state with fashion has been a longterm project of artist Adam Harvey. His camouflage makeup and hair cuts for his CV Dazzle project back in 2012 aimed to throw off the "scent" of facial recognition software by masking your face. Now Harvey is launching a new countersurveillance technique, this time called HyperFace.
The project, a collaboration with interaction studio Hyphen-Labs, features various patterns that are designed to trick facial recognition software into thinking it's seeing eyes, mouths, and the components of a face on the clothing you're wearing. So the pattern can be printed on clothes or textiles and worn round your head or as a scarf or whatever and used to fool the software—software used by the likes of Facebook and retailers—by having it focus on "its face," rather than yours.
It's "overloading [an algorithm] with what it wants," Harvey says. "Oversaturating an area with faces to divert the gaze of the computer vision algorithm," meaning, it can "modify the environment around you, whether it's somebody next to you, whether you're wearing it around your head. "An early prototype fooled the computer vision face detector into seeing over 1,200 possible face detections or 'confidence mappings.'"
HyperFace. Screengrab via
At the moment Harvey has produced a few prototypes, with different patterns targeting different algorithms—neural networks, OpenCV. The project was first revealed in a talk at the Chaos Communications Congress in Hamburg in December 2016 and the final designs are still being developed. But it will launch as a textile print at the Sundance Film Festival on January 16, 2017, as part of Hyphen Labs' NeuroSpeculative AfroFeminism project, which explores black women and their roles in technology and society. The textile prints will also be commercially available, you can sign up to a mailing list to find out when that will be.
HyperFace seems to work as a compliment to the CV Dazzle, although the way it fools facial recognition is decidedly different. "Conceptually, HyperFace recognizes that completely concealing a face to facial detection algorithms remains a technical and aesthetic challenge," notes the project page. "Instead of seeking computer vision anonymity through minimizing the confidence score of a true face (i.e. CV Dazzle), HyperFace offers a higher confidence score for a nearby false face by exploiting a common algorithmic preference for the highest confidence facial region. In other words, if a computer vision algorithm is expecting a face, give it what it wants."
Harvey notes in his Hamburg talk that in 100 years he thinks we'll have a transformation of fashion, one where we will be styling ourselves to optimize our personal privacy and combat mass surveillance. You can watch Harvey's entire talk from the CCC conference, below.