Kyle Kinkade

The idea is surprisingly simple: instead of hiding your face, distort it just enough that a facial-recognition system struggles to understand what it is seeing. Dutch designer Jip van Leeuwenstein created the transparent mask as part of a project called “Surveillance Exclusion” while studying at the Utrecht School of the Arts. The design uses a clear, curved and ribbed surface that bends light passing through it. Facial-recognition systems typically detect landmarks such as the eyes, nose, mouth and jaw, then convert their spatial relationships into a mathematical representation that can be compared with stored faces. The mask changes how those landmarks appear to a camera. Humans can still see much of the wearer’s face and expressions, but the altered geometry can make automated detection and identification more difficult. It is essentially a physical example of adversarial design: changing what a machine sees without completely hiding the underlying object from human perception. However, it isn’t a guaranteed invisibility cloak. Modern recognition systems can be retrained, combine multiple camera angles, and increasingly use other biometric clues such as body shape, gait and movement. The effectiveness would therefore depend heavily on the particular surveillance system. The project was actually created years ago and has recently resurfaced online. Its underlying question has only become more relevant: as cameras become better at identifying people automatically, privacy may increasingly depend on understanding the difference between how humans see us and how machines see us. Source: Jip van Leeuwenstein, Surveillance Exclusion; Utrecht School of the Arts