Astronomers Adapt Galaxy-Measurement Tools to Detect Deepfakes

Researchers at the University of Hull develop a novel method to identify AI-generated images by analyzing light reflections in human eyes.

The technique uses astronomy methods to compare light reflections in the eyeballs of real and AI-generated images.
Inconsistent reflections between a pair of eyes can indicate a deepfake, while consistent reflections suggest authenticity.
The Gini coefficient, a measure of light distribution used in astronomy, proved effective in detecting fake images.
The CAS system, another astronomical tool, was less successful in identifying deepfakes.
This method is not foolproof but offers a new approach in the ongoing effort to detect deepfakes.