Scientists at the University of Buffalo in the United States say that by examining the subject's cornea, one can distinguish deep images made with GAN networks from real images.

Scientists at the University of Buffalo have recently published a paper focusing on a way to identify Deepfake images. According to the Sunnah news agency, 2 scientists say that in this method, the reflection of light from the eye of the subject in the image should be examined. This interesting method is used for hostile grid (GAN) models, which are used to create realistic human faces.

According to scientists at the University of Buffalo, the secret of Deep Fake GAN images lies in the subject's eyes. The published article states that the figures created with the GAN network are used to create fake profiles on social networks and deceive people.

Scientists at the University of Buffalo say figures made with GAN can be identified by "incompatible corneal highlights between the eyes." In comparison, in real images, a kind of reflection is seen in both eyes.