SFFNet: Towards Universal Detection of Generated Images via Spatial-Frequency Domain Fusion

Recent advancements in generative models have enabled the creation of highly realistic images, raising significant concerns about the potential spread of false or deceptive content. Notably, existing methods for AI-generated image detection still face persistent challenges, such as limited accuracy and constrained generalization capabilities. To address these limitations, we propose the Spatial and Frequency Domain Fusion Network (SFFNet), a novel framework that leverages features from both the spatial and frequency domains to enhance detection accuracy and generalization performance. Through extensive cross-generator experiments, we demonstrate that SFFNet significantly outperforms competing detectors, achieving superior average accuracy and demonstrating remarkable generalization capabilities across diverse generative models.