Intel experts have introduced the FakeCatcher neural network, which is claimed to be able to recognize deep fakes in real time with 96% accuracy. This is reported in the PC Mag post.

Intel’s system differs from most analogs in that it looks for signs that are unique to a real person, rather than differences that turn out to be fake. In particular, FakeCatcher, when analyzing the image, records changes in the color of the skin of the face, which is characteristic of the work of the circulatory system of a real person.

“When our heart pumps blood, the veins change color. It is these changes that are captured by the system to determine the authenticity of the image.”– Writes PC Mag.

It is stated that FakeCatcher simultaneously captures indicators from 32 pixels on the face. At the same time, the neural network can detect a fake in a fraction of a second, allowing you to use it to check not only photos but also videos.

Intel believes FakeCatcher could potentially be useful for both social media moderators and journalists to prevent disinformation.