Detecting and Simulating Artifacts in GAN Fake Images

TL;DR

This paper introduces AutoGAN, a GAN simulator that models common artifacts in fake images, and a spectrum-based classifier that detects GAN-generated images effectively without needing real examples from the specific GAN model.

Contribution

The paper presents AutoGAN, a novel GAN simulator for artifact generation, and a spectrum-based classifier that improves detection accuracy without requiring real fake images from the target GAN.

Findings

AutoGAN effectively simulates artifacts across multiple GAN models.

Spectrum-based classifier achieves state-of-the-art detection performance.

Method works without access to real fake images from the target GAN.

Abstract

To detect GAN generated images, conventional supervised machine learning algorithms require collection of a number of real and fake images from the targeted GAN model. However, the specific model used by the attacker is often unavailable. To address this, we propose a GAN simulator, AutoGAN, which can simulate the artifacts produced by the common pipeline shared by several popular GAN models. Additionally, we identify a unique artifact caused by the up-sampling component included in the common GAN pipeline. We show theoretically such artifacts are manifested as replications of spectra in the frequency domain and thus propose a classifier model based on the spectrum input, rather than the pixel input. By using the simulated images to train a spectrum based classifier, even without seeing the fake images produced by the targeted GAN model during training, our approach achieves state-of-the-art performances on detecting fake images generated by popular GAN models such as CycleGAN.