SpecXNet: A Dual-Domain Convolutional Network for Robust Deepfake Detection

2025-09-26

AI Summary

To address insufficient generalization to unseen manipulations and cross-dataset scenarios in deepfake image detection, this paper proposes a Dual-Domain Convolutional Neural Network (Dual-Domain CNN) that jointly models spatial-domain local texture anomalies and frequency-domain global periodic inconsistencies. Methodologically, we introduce a novel Dual-Domain Feature Coupler and Dual Fourier Attention Module to enable content-adaptive cross-domain feature fusion; the end-to-end architecture integrates Fast Fourier Transform (FFT), depthwise separable convolutions, and an enhanced XceptionNet backbone. Evaluated on multiple benchmark datasets, our approach achieves state-of-the-art (SOTA) performance—particularly demonstrating显著 robustness under cross-dataset and previously unseen manipulation types—while maintaining real-time inference capability.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Misinformation & Fake News

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

The increasing realism of content generated by GANs and diffusion models has made deepfake detection significantly more challenging. Existing approaches often focus solely on spatial or frequency-domain features, limiting their generalization to unseen manipulations. We propose the Spectral Cross-Attentional Network (SpecXNet), a dual-domain architecture for robust deepfake detection. The core extbf{Dual-Domain Feature Coupler (DDFC)} decomposes features into a local spatial branch for capturing texture-level anomalies and a global spectral branch that employs Fast Fourier Transform to model periodic inconsistencies. This dual-domain formulation allows SpecXNet to jointly exploit localized detail and global structural coherence, which are critical for distinguishing authentic from manipulated images. We also introduce the extbf{Dual Fourier Attention (DFA)} module, which dynamically fuses spatial and spectral features in a content-aware manner. Built atop a modified XceptionNet backbone, we embed the DDFC and DFA modules within a separable convolution block. Extensive experiments on multiple deepfake benchmarks show that SpecXNet achieves state-of-the-art accuracy, particularly under cross-dataset and unseen manipulation scenarios, while maintaining real-time feasibility. Our results highlight the effectiveness of unified spatial-spectral learning for robust and generalizable deepfake detection. To ensure reproducibility, we released the full code on href{https://github.com/inzamamulDU/SpecXNet}{ extcolor{blue}{ extbf{GitHub}}}.

Problem

• Detecting deepfakes robustly across unseen manipulations
• Integrating spatial and spectral features for detection
• Improving generalization in cross-dataset deepfake scenarios

Innovation

• Dual-domain architecture for robust deepfake detection
• Dual-Domain Feature Coupler captures spatial and spectral features
• Dual Fourier Attention dynamically fuses spatial-spectral features