On Wednesday, research scientists from Facebook who collaborated with the University of Michigan, said that they have found a method to detect deepfake images and trace them in real world settings. This is promising research as this will decrease if not stop the spread of misinformation by people who use deepfake images of celebrities and politicians to spread fake news, which often goes viral.
Two scientists from Facebook— Xi Yin and Tal Hassner wrote that their method will facilitate deepfake detection and would trace it in real world settings, where the deepfake image itself could be the only information available for the detectors to work with.
Deep fakes are altered photos, videos and still images that are used by people with malicious intent to spread fake news and misinformation. People use artificial intelligence (AI) to make this media appear as if it is the real thing. As the technical advances in AI have progressed fake content has become more realistic and sophisticated making it hard for the naked eye to detect the difference between a real and a fake image or video.
Some examples of deepfake videos that went viral include those that used the likeness of actor Tom Cruise, former President Barack Obama, and House Speaker Nancy Pelosi. The videos were realistic, and many believed them till the personalities had to argue and say that the videos were fake.
Facebook is now using the new software it has developed to run through its network to detect deepfake images. The AI program searches for cracks that remain in the manufacturing process that is used by those with malicious intent to change the digital “fingerprint” of an image, just like the unique fingerprints that are left by a digital camera.
The researchers at Facebook hope that their technological progress will be used in the real world to help others fight against fake news used by people who have their own agendas to spread such news. They also hope that their current research could led to greater future technological innovations and investigations.
Facebook said that reverse engineering was used, and the University of Michigan will open source the information including the data, the code, and the training models so that this research is available to a wider number of people and researchers and others could coordinate and investigate disinformation.