Real face foundation representation learning for generalized deepfake detection

Liang Shi, Jie Zhang, Zhilong Ji, Jinfeng Bai, Shiguang Shan

Abstract: Highlights•We improve deepfake detection by modeling the distribution of real faces.•Masked image modeling on real faces reveals potential deepfake artifacts.•Leveraging these exposed artifacts enhances cross-manipulation generalization.•Our method outperforms the state-of-the-art and scales with real face dataset sizes.