Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RCDN: Real-Centered Detection Network for Robust Face Forgery Identification

Wyatt McCurdy, Xin Zhang, Yuqi Song, Min Gao

TL;DR

The Real-Centered Detection Network (RCDN) is proposed, a frequency spatial convolutional neural networks framework with an Xception backbone that anchors its representation space around authentic facial images and reduces the generalization gap compared to leading baselines and achieves the highest cross/in-domain stability ratio.

Abstract

Image forgery has become a critical threat with the rapid proliferation of AI-based generation tools, which make it increasingly easy to synthesize realistic but fraudulent facial content. Existing detection methods achieve near-perfect performance when training and testing are conducted within the same domain, yet their effectiveness deteriorates substantially in crossdomain scenarios. This limitation is problematic, as new forgery techniques continuously emerge and detectors must remain reliable against unseen manipulations. To address this challenge, we propose the Real-Centered Detection Network (RCDN), a frequency spatial convolutional neural networks(CNN) framework with an Xception backbone that anchors its representation space around authentic facial images. Instead of modeling the diverse and evolving patterns of forgeries, RCDN emphasizes the consistency of real images, leveraging a dual-branch architecture and a real centered loss design to enhance robustness under distribution shifts. Extensive experiments on the DiFF dataset, focusing on three representative forgery types (FE, I2I, T2I), demonstrate that RCDN achieves both state-of-the-art in-domain accuracy and significantly stronger cross-domain generalization. Notably, RCDN reduces the generalization gap compared to leading baselines and achieves the highest cross/in-domain stability ratio, highlighting its potential as a practical solution for defending against evolving and unseen image forgery techniques.

RCDN: Real-Centered Detection Network for Robust Face Forgery Identification

TL;DR

The Real-Centered Detection Network (RCDN) is proposed, a frequency spatial convolutional neural networks framework with an Xception backbone that anchors its representation space around authentic facial images and reduces the generalization gap compared to leading baselines and achieves the highest cross/in-domain stability ratio.

Abstract

Image forgery has become a critical threat with the rapid proliferation of AI-based generation tools, which make it increasingly easy to synthesize realistic but fraudulent facial content. Existing detection methods achieve near-perfect performance when training and testing are conducted within the same domain, yet their effectiveness deteriorates substantially in crossdomain scenarios. This limitation is problematic, as new forgery techniques continuously emerge and detectors must remain reliable against unseen manipulations. To address this challenge, we propose the Real-Centered Detection Network (RCDN), a frequency spatial convolutional neural networks(CNN) framework with an Xception backbone that anchors its representation space around authentic facial images. Instead of modeling the diverse and evolving patterns of forgeries, RCDN emphasizes the consistency of real images, leveraging a dual-branch architecture and a real centered loss design to enhance robustness under distribution shifts. Extensive experiments on the DiFF dataset, focusing on three representative forgery types (FE, I2I, T2I), demonstrate that RCDN achieves both state-of-the-art in-domain accuracy and significantly stronger cross-domain generalization. Notably, RCDN reduces the generalization gap compared to leading baselines and achieves the highest cross/in-domain stability ratio, highlighting its potential as a practical solution for defending against evolving and unseen image forgery techniques.
Paper Structure (19 sections, 11 equations, 2 figures, 3 tables)

This paper contains 19 sections, 11 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Diffusion-based facial forgery benchmarks
  4. Cross-domain generalization
  5. Proposed Methods
  6. Model Architecture
  7. Spatial branch
  8. Frequency branch
  9. Feature fusion and projection
  10. Supervision heads
  11. Loss Function Design
  12. Classification loss
  13. Real-centered loss
  14. Separation loss
  15. Experiments
  16. ...and 4 more sections

Figures (2)

  • Figure 1: Examples of diffusion-based facial forgery categories considered in this work: Face Editing (FE) alters specific attributes of a real face, Image-to-Image translation (I2I) refines or transforms facial appearance from an input image, and Text-to-Image generation (T2I) creates a new face guided purely by textual descriptions.
  • Figure 2: Overview of the proposed Real-Centered Detection Network (RCDN). The model processes an input face image through a frequency branch (FFT-based ConvNet) and a spatial branch (Xception backbone), concatenates the resulting features, and projects them into a 128-D embedding space. This embedding is supervised both by a classifier and by real-centered constraints. By anchoring the feature space around real images, RCDN achieves robustness across forgery domains.