Stories

Fooling facial recognition with fashion

Illustrated collage of a woman wearing an anti-tan mask.
Illustration: Aïda Amer/Axios

Some designers, researchers and activists are trying to fool facial recognition technologies with fashion.

What’s happening: The protests in Hong Kong have drummed up new interest in anti-surveillance fashion, according to designers Adam Harvey and Scott Urban.

The big picture: Data repositories contain millions of faces compiled everywhere from social networks to dating sites to cameras at restaurants — with no oversight of how they’re collected.

From sunglasses to hoodies, adversarial fashions are becoming increasingly popular as facial recognition use spreads — despite calls to ban the technology.

  • Privacy eyewear: Urban designs IRpair sunglasses, which prevent even your iPhone from recognizing you — and dupe infrared facial recognition technologies.
  • Camouflage makeup: Computer Vision Dazzle is an anti-surveillance makeup project by Harvey, which tricks facial recognition algorithms by using unusual makeup tones, concealing the nose and creating asymmetry. But the flashy makeup can make you more visible to other humans.
  • False-face patterns: Harvey's Hyperface is a textile pattern that fools facial recognition software into detecting facial parts like eyes, mouths and noses on your clothing to focus on these “faces” rather than yours.
  • Anti-paparazzi devices: CamoFlash, also by Harvey, triggers LEDs that overexpose paparazzi camera flashes.
  • Anti-drone hoodies and hijabs contain shiny fabrics that reflect thermal radiation, to “avert overhead thermal surveillance," designed by Harvey.
  • Patches on beanie hats designed by researchers from Moscow trick the state-of-the-art facial recognition system ArcFace.
  • Minimalist brass masks by Polish designer Ewa Nowak, called Incognito, deflect facial recognition software like the DeepFace algorithm used by Facebook.

But, but, but: Academics like Torin Monahan argue that designs like CV Dazzle normalize the surveillance state and contribute to the “harmful conditions they seek to change.”

At stake is the question of what is considered identity in this new era of surveillance, where travelers crossing the U.S. border are subject to facial biometrics collection and truck drivers are monitored with infrared facial recognition systems.

  • Azeem Azhar writes in his newsletter, Exponential View, "Is our future one where we have to disguise our real selves (and emotional states) in order to be true to them?"

What's next: There will never be a way to pull off of every system of surveillance, Urban says, even with innovations like privacy eyewear and clothing.

  • But Harvey disagrees. The point of his designs are "to show people that surveillance has its vulnerabilities."
  • "The challenge now," he says, "is looking at how people are looking at you."

Go deeper: