Astronomers develop technology to detect deep fakes through eye reflections – IT Pro – News

Astronomers at the University of Hull in the UK have developed a technique to detect deep fakes using eye reflections. The technology is based on instruments that astronomers use to study galaxies.

The technology was presented last week at a meeting of the British Royal Astronomical Society. To detect whether a photo is a deep fake, the reflections of both eyes are analyzed. The shape, location and intensity of light reflections are taken into account. During the analysis, the researchers apply techniques normally used in astronomy to measure the distribution of light in galaxy images.

In real photographs, the light reflections in both eyes are usually uniform because they are caused by the same light sources. In deepfakes, reflections can be inconsistent, as AI models often struggle to reproduce light reflections accurately. While this method can be effective at detecting deep fakes, it can also produce false positives, according to astronomers. “This method gives us a foundation, a plan of attack in an arms race to detect deep fakes,” Kevin Pimblett of the University of Hull told the Royal Astronomical Society.