Deepfakes Can Be Detected by Borrowing a Method From Astronomy

Close-up of a human eye with the iris replaced by a vibrant cosmic scene. The iris appears to be filled with galaxies, stars, and nebulae, creating a stunning and surreal visual effect. The detailed eyelashes and skin texture are also visible.

AI images and deepfakes can be detected by analyzing a person’s eyes by borrowing a technique from astronomy.

According to new research shared by the Royal Astronomical Society, AI-generated fakes can be analyzed the same way astronomers study galaxies.

University of Hull MSc student Adejumoke Owolabi concludes that it is all about the reflection in a person’s eyes. If the reflections match, the image is likely to be that of a real human. If they don’t, they’re probably deepfakes.

“The reflections in the eyeballs are consistent for the real person, but incorrect (from a physics point of view) for the fake person,” explains Kevin Pimbblet, professor of astrophysics and director of the Centre of Excellence for Data Science, Artificial Intelligence, and Modelling at the University of Hull.

Split image showing a woman with long brown hair and a smiling expression on the left, and a man with curly black hair and a neutral expression on the right. Below each image are three close-up images of various eye expressions.
Real person Scarlett Johansson, left, has matching reflections in her eyes. The AI-generated person on the right has non-matching reflections.

The researchers analyzed reflections of light on the eyeballs of people in real and AI-generated images. They then borrowed a method typically used in astronomy to quantify the reflections and checked for consistency between left and right eyeball reflections.

“To measure the shapes of galaxies, we analyze whether they’re centrally compact, whether they’re symmetric, and how smooth they are. We analyse the light distribution,” says Professor Pimbblet.

“We detect the reflections in an automated way and run their morphological features through the CAS [concentration, asymmetry, smoothness] and Gini index to compare similarity between left and right eyeballs.

“The findings show that deepfakes have some differences between the pair.”

The image displays rows of close-up shots of human eyes. Each row shows a pair of eyes, with the right side of the row containing the same eyes but with blue circles and green and red markings over the pupils and irises, indicating some form of analysis or tracking.
A series of eyeballs belonging to deepfakes.

The Gini index is normally used to measure how the light in an image of a galaxy is distributed among its pixels. This measurement is made by ordering the pixels that make up the image of a galaxy in ascending order by flux and then comparing the result to what would be expected from a perfectly even flux distribution.

A Gini value of 0 is a galaxy in which the light is evenly distributed across all of the image’s pixels, while a Gini value of 1 is a galaxy with all light concentrated in a single pixel.

A series of close-up photos of four different sets of eyes in two rows. The left column shows original eye images, while the right column highlights detection points around the eyes, with circles and dots of various colors indicating areas of focus or interest.
A series of real eyeballs with matching reflections.

“It’s important to note that this is not a silver bullet for detecting fake images,” Professor Pimbblet says.

“There are false positives and false negatives; it’s not going to get everything. But this method provides us with a basis, a plan of attack, in the arms race to detect deepfakes.”


Image credits: Royal Astronomical Society.

FOLLOW US ON GOOGLE NEWS

Read original article here

Denial of responsibility! Swift Telecast is an automatic aggregator of the all world’s media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, all materials to their authors. If you are the owner of the content and do not want us to publish your materials, please contact us by email – swifttelecast.com. The content will be deleted within 24 hours.

Leave a Comment