CrossDF: improving cross-domain deepfake detection with deep information decomposition
Frontiers in Big Data · 18.11.2025 · 8
DOI: 10.3389/fdata.2025.1669488 ↗Kurz gefasst
Fachartikel in der Zeitschrift Frontiers in Big Data, online erschienen am 18.11.2025. Verfasst von Shanmin Yang und 7 weiteren Autor:innen. Beteiligte Einrichtung: Computer Science and Technology, Chengdu University of Information Technology, Chengdu, China. Die Arbeit wurde bisher 12-mal zitiert (Crossref) und nennt 54 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Deepfake technology represents a serious risk to safety and public confidence. While current detection approaches perform well in identifying manipulations within datasets that utilize identical deepfake methods for both training and validation, they experience notable declines in accuracy when applied to cross-dataset situations, where unfamiliar deepfake techniques are encountered during testing. To tackle this issue, we propose a Deep Information Decomposition (DID) framework to improve Cross-dataset Deepfake Detection (CrossDF). Distinct from most existing deepfake detection approaches, our framework emphasizes high-level semantic attributes instead of focusing on particular visual anomalies. More specifically, it intrinsically decomposes facial representations into deepfake-relevant and unrelated components, leveraging only the deepfake-relevant features for classification between genuine and fabricated images. Furthermore, we introduce an adversarial mutual information minimization strategy that enhances the separability between these two types of information through decorrelation learning. This significantly improves the model's robustness to irrelevant variations and strengthens its generalization capability to previously unseen manipulation techniques. Extensive experiments demonstrate the effectiveness and superiority of our proposed DID framework for cross-dataset deepfake detection. It achieves an AUC of 0.779 in cross-dataset evaluation from FF++ to CDF2 and improves the state-of-the-art AUC significantly from 0.669 to 0.802 on the diffusion-based Text-to-Image dataset.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Computer Science and Technology, Chengdu University of Information Technology, Chengdu, China.
- University at Buffalo, State University of New York (SUNY), Buffalo, NY, United States.
- Purdue University, West Lafayette, IN, United States.
- Microsoft Research Asia, Beijing, China.
- University at Albany, State University of New York (SUNY), Albany, NY, United States.
Zitieren
APA
Yang, S., Guo, H., Hu, S., Zhu, B., Fu, Y., Lyu, S., Wu, X., & Wang, X. (2025). CrossDF: improving cross-domain deepfake detection with deep information decomposition. Frontiers in Big Data, 8. https://doi.org/10.3389/fdata.2025.1669488
Vancouver
Yang S, Guo H, Hu S, Zhu B, Fu Y, Lyu S, et al. CrossDF: improving cross-domain deepfake detection with deep information decomposition. Frontiers in Big Data. 2025;8. doi:10.3389/fdata.2025.1669488
BibTeX
@article{Yang2025,
author = {Yang, Shanmin and Guo, Hui and Hu, Shu and Zhu, Bin and Fu, Ying and Lyu, Siwei and Wu, Xi and Wang, Xin},
title = {CrossDF: improving cross-domain deepfake detection with deep information decomposition},
journal = {Frontiers in Big Data},
year = {2025},
volume = {8},
doi = {10.3389/fdata.2025.1669488}
}
EUVIMED