Are Benford's Law-Based Detectors Effective for GAI Generated Images?

Benford's Law (BL) has been proposed as a method to distinguish natural images from those that were altered in some way, through manipulation or distortion, after their acquisition. Another common use of BL is to detect artificially generated images. In this article, we evaluate the effectiveness of BL in the detection of Generative Artificial Intelligence (GAI) generated images. The images were generated in four different categories depending on the prompts to the GAI, namely original, artificial, cartoon-like and realistic. The results showed that only about 60% of the artificial images do not fit BL, while the other 40 % do. However, some very artificial looking images pass the threshold of BL. We hope that this article sheds some light on the topic of using BL for image forensics.