Wavelet-based GAN Fingerprint Detection using ResNet50

2025-10-21

AI Summary

To address the challenge of detecting GAN-generated images, this paper proposes a wavelet-domain-based deepfake detection method. We apply discrete wavelet transform (DWT) using both Haar and Daubechies wavelets to perform multi-scale frequency decomposition of input images, explicitly modeling high-frequency artifacts—termed “fingerprints”—characteristic of StyleGAN outputs in the wavelet coefficients. The resulting wavelet subband images are fed directly into a ResNet50 backbone for end-to-end binary classification. Experiments demonstrate substantial improvements over purely spatial-domain baselines: the Daubechies-based model achieves 95.1% accuracy, while the Haar-based variant attains 93.8%, both significantly surpassing the spatial-domain baseline’s 81.5%. This work provides the first systematic empirical validation of the discriminative power of wavelet-domain features for GAN image attribution, establishing a novel frequency-aware paradigm for digital forensic analysis.

Technology Category

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

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Identifying images generated by Generative Adversarial Networks (GANs) has become a significant challenge in digital image forensics. This research presents a wavelet-based detection method that uses discrete wavelet transform (DWT) preprocessing and a ResNet50 classification layer to differentiate the StyleGAN-generated images from real ones. Haar and Daubechies wavelet filters are applied to convert the input images into multi-resolution representations, which will then be fed to a ResNet50 network for classification, capitalizing on subtle artifacts left by the generative process. Moreover, the wavelet-based models are compared to an identical ResNet50 model trained on spatial data. The Haar and Daubechies preprocessed models achieved a greater accuracy of 93.8 percent and 95.1 percent, much higher than the model developed in the spatial domain (accuracy rate of 81.5 percent). The Daubechies-based model outperforms Haar, showing that adding layers of descriptive frequency patterns can lead to even greater distinguishing power. These results indicate that the GAN-generated images have unique wavelet-domain artifacts or "fingerprints." The method proposed illustrates the effectiveness of wavelet-domain analysis to detect GAN images and emphasizes the potential of further developing the capabilities of future deepfake detection systems.

Problem

• Detecting GAN-generated images using wavelet transforms and ResNet50
• Differentiating StyleGAN images from real ones via frequency artifacts
• Improving detection accuracy with wavelet preprocessing over spatial methods

Innovation

• Uses wavelet transform preprocessing for image analysis
• Employs ResNet50 network for classification of images
• Detects GAN fingerprints in frequency domain artifacts