Facebook scientists have developed AI software that detects deepfakes and also identifies who is distributing them.
Facebook researchers Tal Hassner and Xi Yin said their team has worked with the University of Michigan to create software that detects deepfakes in order to figure out how they were made and where they are distributed from.
Deepfake is a realistic manipulation of audio and video materials using artificial intelligence.
Our method will make deepfake detection and tracking easier. Usually, the program only has to work with the image, there is no other information.
The developers note that any correction of images leaves “prints” – these are unique patterns left on the images: the new software works according to these principles.
The algorithm will help to efficiently investigate incidents with deepfakes, the authors note.