On Wednesday, Facebook researchers Xi Yin and Tal Hassner revealed they have developed artificial intelligence that can identify so-called “deepfakes” and trace their origins through reverse engineering.
While the work was done in collaboration with Michigan State University, the new Facebook software guides deepfake images through its network, and the company’s AI program searches for cracks left in the manufacturing process to alter an image’s digital “fingerprints.”
According to the Facebook researchers, “Our method will facilitate deepfake detection and tracing in real-world settings, where the deepfake image itself is often the only information detectors have to work with.”
For more information, read the original story in NPR.