Fourier-Based GAN Fingerprint Detection using ResNet50

2025-10-18

AI Summary

To address the challenge of distinguishing StyleGAN-generated images from authentic ones, this paper proposes an interpretable image forensics method that synergistically integrates frequency-domain analysis with deep learning. The core contribution lies in identifying and leveraging a distinctive periodic artifact—ubiquitous in GAN-generated images within the frequency domain—as a discriminative fingerprint. Specifically, input images are transformed into the frequency domain via 2D discrete Fourier transform (DFT), and the resulting spectral representations are classified end-to-end using ResNet-50. Compared to conventional spatial-domain models, the proposed frequency-aware approach achieves substantial performance gains: 92.8% classification accuracy and an AUC of 0.95. Crucially, the frequency-domain features exhibit strong interpretability, enabling visual inspection and diagnostic analysis of generation artifacts. This work establishes a novel paradigm for synthetic image detection and delivers a practical, principled tool for digital media authentication.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

The rapid rise of photorealistic images produced from Generative Adversarial Networks (GANs) poses a serious challenge for image forensics and industrial systems requiring reliable content authenticity. This paper uses frequency-domain analysis combined with deep learning to solve the problem of distinguishing StyleGAN-generated images from real ones. Specifically, a two-dimensional Discrete Fourier Transform (2D DFT) was applied to transform images into the Fourier domain, where subtle periodic artifacts become detectable. A ResNet50 neural network is trained on these transformed images to differentiate between real and synthetic ones. The experiments demonstrate that the frequency-domain model achieves a 92.8 percent and an AUC of 0.95, significantly outperforming the equivalent model trained on raw spatial-domain images. These results indicate that the GAN-generated images have unique frequency-domain signatures or "fingerprints". The method proposed highlights the industrial potential of combining signal processing techniques and deep learning to enhance digital forensics and strengthen the trustworthiness of industrial AI systems.

Problem

• Detecting GAN-generated images using frequency-domain analysis
• Identifying unique Fourier fingerprints in synthetic images
• Improving digital forensics with deep learning and signal processing

Innovation

• Using Fourier transform to detect GAN fingerprints
• Training ResNet50 on frequency-domain images
• Combining signal processing with deep learning