This startup tricks facial recognition algorithms with clothes
Rachele Didero is a 26-year-old Italian engineer and fashion designer. Committed to the protection of privacy, she developed as part of her doctorate at the Polytechnic of Milan, patterns capable of deceiving Yolo, the most widespread face recognition algorithm.
An AI to trick facial recognition
In order to define the pattern of the garment intended to fool facial recognition software, Cap_able uses an artificial intelligence that generates a pattern called “conflicting patch”. The latter aims to cover their tracks by pretending to be an animal. Facial recognition is primarily responsible for identifying the individual, and if it is a human, collects biometric data from the face. Confronted with what the camera thinks is a dog, a zebra or a giraffe, Yolo cannot therefore classify the face as human and allows the wearer to keep their biometric data.
However, a major drawback remains. While Yolo can benefit from constant updates, the clothes themselves cannot change patterns. It would thus be necessary to continually renew its wardrobe to preserve its anonymity.
The collection is already available for pre-order on the brand’s website. Count 277 € for trousers and 390 € for a sweater. More than equipping the population en masse, the objective is more to warn against the dangers that technologies that collect personal information without your knowledge can represent.