Generalized Deepfake Attribution

TL;DR

This paper introduces a generalized deepfake attribution network capable of identifying the originating GAN architecture of fake images, even when models are retrained with different seeds or fine-tuned, addressing seed dependency issues.

Contribution

The paper presents a novel GDA-Net that effectively attributes deepfakes to their source GANs across different seeds and fine-tuned models, improving robustness over existing methods.

Findings

High accuracy in cross-seed attribution tasks

Effective attribution for fine-tuned GAN models

Outperforms existing attribution methods

Abstract

The landscape of fake media creation changed with the introduction of Generative Adversarial Networks (GAN s). Fake media creation has been on the rise with the rapid advances in generation technology, leading to new challenges in Detecting fake media. A fundamental characteristic of GAN s is their sensitivity to parameter initialization, known as seeds. Each distinct seed utilized during training leads to the creation of unique model instances, resulting in divergent image outputs despite employing the same architecture. This means that even if we have one GAN architecture, it can produce countless variations of GAN models depending on the seed used. Existing methods for attributing deepfakes work well only if they have seen the specific GAN model during training. If the GAN architectures are retrained with a different seed, these methods struggle to attribute the fakes. This seed dependency issue made it difficult to attribute deepfakes with existing methods. We proposed a generalized deepfake attribution network (GDA-N et) to attribute fake images to their respective GAN architectures, even if they are generated from a retrained version of the GAN architecture with a different seed (cross-seed) or from the fine-tuned version of the existing GAN model. Extensive experiments on cross-seed and fine-tuned data of GAN models show that our method is highly effective compared to existing methods. We have provided the source code to validate our results.