Logo image
Back
Making Generated Images Hard to Spot: A Transferable Attack on Synthetic Image Detectors
Book chapter   Open access

Making Generated Images Hard to Spot: A Transferable Attack on Synthetic Image Detectors

Xinwei Zhao and Matthew C. Stamm
Pattern Recognition, Computer Vision, and Image Processing. ICPR 2022 International Workshops and Challenges
01 Jan 2023
DOI: https://doi.org/10.1007/978-3-031-37745-7_6
url
https://arxiv.org/abs/2104.12069View
SubmittedOpen Access (License Unspecified),  Open

Abstract

Anti-forensics Forensic detectors GAN-based attacks
Visually realistic GAN-generated images have recently emerged as an important misinformation threat. Research has shown that these synthetic images contain forensic traces that are readily identifiable by forensic detectors. Unfortunately, these detectors are built upon neural networks, which are vulnerable to recently developed adversarial attacks. In this paper, we propose a new anti-forensic attack capable of fooling GAN-generated image detectors. Our attack uses an adversarially trained generator to synthesize traces that these detectors associate with real images. Furthermore, we propose a technique to train our attack so that it can achieve transferability, i.e. it can fool unknown CNNs that it was not explicitly trained against. We evaluate our attack through an extensive set of experiments, where we show that our attack can fool eight state-of-the-art detection CNNs with synthetic images created using seven different GANs, and outperform other alternative attacks.

Metrics

12 Record Views
2 citations in Scopus

Details

Logo image