Inventing a tool that will spot deepfakes 94 percent of the time

Published March 15, 2021

Print

CNET reported on a new UB research study led by Siwei Lyu, that describes a tool that spots deepfake photos by examining light reflection in the eyes. The tool is 94% effective with portrait photos. “The two eyes should have very similar reflective patterns because they’re seeing the same thing," he said. "It’s something that we typically don’t typically notice when we look at a face."

 

Media Contact Information

Media Relations (University Communications)
330 Crofts Hall (North Campus)
Buffalo, NY 14260-7015
Tel: 716-645-6969
ub-news@buffalo.edu