Facebook Develops AI to Bypass Facial Recognition Systems

Artificial intelligence experts at Facebook have developed a real-time “de-identification” system for users. According to early testing, this AI is capable of bypassing modern facial recognition systems. The method slightly distorts a person’s face in a way that confuses facial recognition software, while still allowing other people to recognize the individual.

“Facial recognition can lead to a loss of privacy, and this face-swapping technology could be used to create misleading videos. Recent events involving the advancement and misuse of facial recognition technology highlight the need to understand methods that can successfully achieve de-identification,” states the document describing the technology.

Facebook’s approach combines a generative adversarial autoencoder and a network of classifiers. Similar to Deepfake face-swapping software, the AI uses an encoder-decoder architecture to generate both a mask and an image. The system produces both distorted and undistorted images of a person’s face, which can be embedded into videos.

Currently, Facebook does not plan to implement this technology in any of its applications. The company intends to present the method at the International Conference on Computer Vision, held in Seoul, South Korea.

Source

• Our other channels
• Our friends and partners