Deepfake images of people are used largely by bots to spread fake news and sway opinion. While they look real, everything about them is synthetic — just a series of codes that come together to form an image of a person who doesn’t exist. This global and interdisciplinary issue begs the question: Is seeing really believing?
As Dr. Freddie Witherden and doctoral student Tarik Dzanic discovered, the falsified faces are not without fault.
Witherden explained real photographs have a noise pattern associated with their pixels. The human eye cannot notice this noise, so the bots that generate deepfake images do not pay attention to that detail when creating the images. They only focus on the big picture and recognizable features of a face, such as a mouth, nose and eyes. Witherden’s algorithm detects this noise to identify deepfake images.
“(The bots) are not concerned about the minutia associated with the noise and, so, they don’t make any effort to try to get that noise profile correct,” said Witherden. “By doing some very simple signal processing, we can identify these discrepancies and thus discriminate between real or fake images.”
Witherden was named in the 2020 Forbes’ 30 Under 30 list for Europe in the science and health care category.
“We are delighted, but not surprised, that Dr. Witherden was mentioned in this prestigious list,” said department head Dr. Sharath Girimaji. “It is a fitting recognition of his incredible talent and hard work. We have an excellent group of young faculty members, and I look forward to exciting times ahead for the ocean engineering department.”