Deepfake Image Testing Tool Has Been Designed By the University Of Buffalo Research Team
The researchers have designed a tool that is going to help compare an original and a deepfake image of people. The results obtained from this tool are close to 94% correct and effective. The tool focuses on the reflection of the person’s eyes to say whether what it is watching is genuine or a deepfake one. Researchers claim that such images have reflections that are unmatched.
These researchers belong to the University of Buffalo, and this has even been tried on the CEO of Facebook, Mark Zuckerberg.
According to the research papers, the people who have developed it has built a process of spotting a picture. They will look closely into the eyes of the image and then give their results accordingly. The researchers have found that reflections in the cornea of a person’s eye are identical to that of any other person. Whereas in the case of deepfaked images, the case was different. During the testing phase, the researchers claimed that the technology has become 94% effective.
Cons of the Deepfake Software
However, there are a few limitations as well- it becomes a bit less effective on pictures that are not clicked in portrait mode. In such photos, there are significantly fewer chances to get proper reflections in people’s eyes.
Deepfake images are those which have been digitally generated or designed using some kind of artificial intelligence technology. This particular kind of technology is called Generative Adversarial Networks (GANs). Such software can also be used to swap faces on videos with somebody else.