Scientists working at Facebook say they have developed artificial intelligence (AI) software that can not only recognize used images using the “deepfake” trick, but also trace the source of the distribution.
“Deepfake” is a technology that uses AI to create fake content, be it images, videos, even audio, that are difficult to distinguish in the usual way. Therefore, this trick is often used to create fake news and malicious scams.
Two scientists Tal Hassner and Xi Yin and their colleagues at Facebook have collaborated with the University of Michigan (USA) to create software that can recover images edited with “deepfake” technology. can identify the source of these images.
The method of tracing is based on finding and detecting imperfections in photos or clips during a “deepfake” intervention that scientists call a way to change a technical “fingerprint”. number of images.
In the announcement, the scientists stated: “Our method will facilitate the detection and tracking of deepfake images in the real world.”
The experts further noted: “This work will provide a better tool to help researchers and professionals investigate cases of information counterfeiting using deepfake technology, as well as open up new avenues new research in the future”.
Late last year, technology group Microsoft announced software to detect images or clips edited with “deepfake” technology through Video Authenticator software, during the US presidential election.