Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Wavelet-based GAN Fingerprint Detection using ResNet50

Sai Teja Erukude, Suhasnadh Reddy Veluru, Viswa Chaitanya Marella

TL;DR

The paper addresses the challenge of detecting GAN-generated images by leveraging wavelet-domain preprocessing to expose frequency-domain fingerprints. It introduces a pipeline that applies 2D discrete wavelet transforms using Haar and Daubechies wavelets to produce multi-resolution representations, which are then classified by a ResNet50 network, and it benchmarks against a spatial-domain baseline. Empirically, wavelet-based inputs substantially improve detection performance, with Daubechies-2 achieving up to 95.1% accuracy and an AUC of 0.97, outperforming the spatial model at 81.5% accuracy. The findings demonstrate the value of frequency-domain analysis in deepfake detection and point toward robust, multi-domain forensic systems that can adapt to evolving GANs.

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.

Wavelet-based GAN Fingerprint Detection using ResNet50

TL;DR

The paper addresses the challenge of detecting GAN-generated images by leveraging wavelet-domain preprocessing to expose frequency-domain fingerprints. It introduces a pipeline that applies 2D discrete wavelet transforms using Haar and Daubechies wavelets to produce multi-resolution representations, which are then classified by a ResNet50 network, and it benchmarks against a spatial-domain baseline. Empirically, wavelet-based inputs substantially improve detection performance, with Daubechies-2 achieving up to 95.1% accuracy and an AUC of 0.97, outperforming the spatial model at 81.5% accuracy. The findings demonstrate the value of frequency-domain analysis in deepfake detection and point toward robust, multi-domain forensic systems that can adapt to evolving GANs.

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.
Paper Structure (13 sections, 10 figures, 2 tables)

This paper contains 13 sections, 10 figures, 2 tables.

Table of Contents

  1. Introduction and Related Work
  2. Data and CNN Architecture
  3. Real Images Collection
  4. GAN-generated Images Collection
  5. Dataset Composition
  6. CNN Architecture
  7. Methodology
  8. Wavelet Transform
  9. Haar Wavelet
  10. Daubechies Wavelet
  11. Training Setup
  12. Results and Discussion
  13. Conclusion

Figures (10)

  • Figure 1: A sample of real images, including human faces (from FFHQ) and cat images (from the Cats vs. Dogs dataset). These images are the true class for the binary classification task. They also provide variety in terms of age, appearance, and background.
  • Figure 2: Sample StyleGAN2-generated (fake) images, including synthetic human faces and images of cats. While they are visually realistic, these images contain artifacts (GAN fingerprints) that deep learning models can utilize in the wavelet domain for detection.
  • Figure 3: High-level architecture of the ResNet50 deep neural network used for both spatial-domain and wavelet-domain classification tasks.
  • Figure 4: Illustration of the Haar and Daubechies wavelet filters used for image decomposition, highlighting differences in filter shapes.
  • Figure 5: Example of a real cat image before and after applying Haar wavelet.
  • ...and 5 more figures