A Spatial-Frequency Aware Multi-Scale Fusion Network for Real-Time Deepfake Detection

With the rapid advancement of real-time deepfake generation techniques, forged content is becoming increasingly realistic and widespread across applications like video conferencing and social media. Although state-of-the-art detectors achieve high accuracy on standard benchmarks, their heavy computational cost hinders real-time deployment in practical applications. To address this, we propose the Spatial-Frequency Aware Multi-Scale Fusion Network (SFMFNet), a lightweight yet effective architecture for real-time deepfake detection. We design a spatial-frequency hybrid aware module that jointly leverages spatial textures and frequency artifacts through a gated mechanism, enhancing sensitivity to subtle manipulations. A token-selective cross attention mechanism enables efficient multi-level feature interaction, while a residual-enhanced blur pooling structure helps retain key semantic cues during downsampling. Experiments on several benchmark datasets show that SFMFNet achieves a favorable balance between accuracy and efficiency, with strong generalization and practical value for real-time applications.

Key Contributions

• Spatial-frequency hybrid aware module fusing wavelet features and coordinate attention via a dynamic gating map to enhance forgery region perception
• Token-selective cross attention module for efficient cross-scale feature interaction and forgery feature alignment
• Residual downsampling module based on blur pooling to preserve structural and edge details while reducing aliasing

🛡️ Threat Analysis

Details

Similar Papers