Attributing Fake Images to GANs: Learning and Analyzing GAN Fingerprints

TL;DR

This paper investigates how GANs leave unique, stable fingerprints in generated images, enabling attribution and source identification, even under adversarial conditions, advancing digital forensics and model authentication.

Contribution

It introduces the first method to learn and analyze GAN fingerprints for image attribution and source identification, demonstrating their stability and robustness.

Findings

GANs leave distinct, stable fingerprints in generated images.

Minor training differences produce different fingerprints, enabling fine-grained authentication.

Fingerprints persist across image frequencies and are resilient to adversarial perturbations.

Abstract

Recent advances in Generative Adversarial Networks (GANs) have shown increasing success in generating photorealistic images. But they also raise challenges to visual forensics and model attribution. We present the first study of learning GAN fingerprints towards image attribution and using them to classify an image as real or GAN-generated. For GAN-generated images, we further identify their sources. Our experiments show that (1) GANs carry distinct model fingerprints and leave stable fingerprints in their generated images, which support image attribution; (2) even minor differences in GAN training can result in different fingerprints, which enables fine-grained model authentication; (3) fingerprints persist across different image frequencies and patches and are not biased by GAN artifacts; (4) fingerprint finetuning is effective in immunizing against five types of adversarial image perturbations; and (5) comparisons also show our learned fingerprints consistently outperform several baselines in a variety of setups.