Towards Discovery and Attribution of Open-world GAN Generated Images

TL;DR

This paper introduces an iterative method for detecting and attributing images to both known and unseen GANs, addressing the challenge of open-world scenarios where new GANs continuously emerge.

Contribution

The authors propose a novel algorithm that discovers and attributes images to previously unseen GANs by leveraging their unique fingerprints, extending beyond closed-set limitations.

Findings

High accuracy in discovering unseen GANs

Effective generalization to GANs trained on new datasets

Applicable to real/fake detection in open-world settings

Abstract

With the recent progress in Generative Adversarial Networks (GANs), it is imperative for media and visual forensics to develop detectors which can identify and attribute images to the model generating them. Existing works have shown to attribute images to their corresponding GAN sources with high accuracy. However, these works are limited to a closed set scenario, failing to generalize to GANs unseen during train time and are therefore, not scalable with a steady influx of new GANs. We present an iterative algorithm for discovering images generated from previously unseen GANs by exploiting the fact that all GANs leave distinct fingerprints on their generated images. Our algorithm consists of multiple components including network training, out-of-distribution detection, clustering, merge and refine steps. Through extensive experiments, we show that our algorithm discovers unseen GANs with high accuracy and also generalizes to GANs trained on unseen real datasets. We additionally apply our algorithm to attribution and discovery of GANs in an online fashion as well as to the more standard task of real/fake detection. Our experiments demonstrate the effectiveness of our approach to discover new GANs and can be used in an open-world setup.