Facebook’s AI research lab has apparently developed a system that could protect an individual’s identity on video by generating a fake face.
This ‘de-identification’ technology works much like the controversial deepfake tech that has been used to create fake news videos and to morph the faces of celebrities on to that of porn stars. In this case, Facebook is using the technology to morph a person’s face so that they remain unrecognisable to facial recognition technology.
According to Venturebeat, who spoke to Facebook AI Research Engineer Lior Wolf, the scope of the technology can be expanded to include the morphing of a person’s voice, mannerisms, online behaviour, “or any other type of identifiable information that you want to remove.”
As with deepfake tech, the system uses samples of the person’s video and distorts the images. These distorted images are then compared to the original images and a suitable, natural-looking mask is generated. This mask can be superimposed on videos.