This new tool can identify deepfake photos through tiny reflections in the eyes

In recent years, artificial intelligence advances have given rise to a new type of manipulated video called “deepfakes,” which, according to reports in other countries’ media, has increased the spread of misinformation. A tool to help distinguish between authentic and authentic edits is needed as these edits become more convincing. Researchers at the University of Buffalo have come up with a way to accomplish this.

By analyzing tiny reflections in the eyes, this new technology claims to be able to detect deep forgery with a 94% accuracy rate.

A deep learning algorithm is trained on a real video clip of a person and then mixed with computer images to create a fictitious video clip as a form of artificial media.

Increasingly, these things are becoming a fact of everyday life.

There are growing concerns about the potential impact of this technology on democracy from experts. It appears that politicians, such as Nancy Pere, are saying things they have never said before in their tampered video. The video of Rosie’s conversation has been tampered with, according to her.

To develop a new deep forgery detection tool by exploiting tiny deviations in eye reflection, the University of Buffalo team was led by computer scientist Siwei Lyu.

When people look at something in real life, they see the same shape and color in the object they are looking at.

Thus, any light that enters the eye from the light source is imaged on the cornea in some reflective mode in both eyes,” said Lyu. “The end corneal almost like a beautiful semicircular and has a strong reflecting capacity.” Due to the fact that they are both looking at exactly the same thing, the results should be very similar. When we look at someone’s face, we don’t usually notice this. “

Deep forgery, on the other hand, is a different story.

Lyu and his team created a set of computer tools to address this issue. They begin by sketching the face, then move on to the eyes, eyeballs, and finally the refraction of light on each one. The tool is able to distinguish between slight variations in the reflected light’s shape, color, and intensity after performing a detailed comparison.

According to researchers, a new tool they developed can distinguish between real people and deep-fake images with 94% accuracy. This number is encouraging, but the research team cautioned that this method has some drawbacks. Editing software can fix some of these issues; however, the detected image must show a clear view of the eye.


Please enter your comment!
Please enter your name here