{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,21]],"date-time":"2026-05-21T17:20:35Z","timestamp":1779384035075,"version":"3.53.1"},"reference-count":54,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,11,18]],"date-time":"2025-11-18T00:00:00Z","timestamp":1763424000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Big Data"],"abstract":"<jats:p>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.<\/jats:p>","DOI":"10.3389\/fdata.2025.1669488","type":"journal-article","created":{"date-parts":[[2025,11,19]],"date-time":"2025-11-19T21:34:39Z","timestamp":1763588079000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":2,"title":["CrossDF: improving cross-domain deepfake detection with deep information decomposition"],"prefix":"10.3389","volume":"8","author":[{"given":"Shanmin","family":"Yang","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hui","family":"Guo","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shu","family":"Hu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bin","family":"Zhu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ying","family":"Fu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Siwei","family":"Lyu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xi","family":"Wu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xin","family":"Wang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2025,11,18]]},"reference":[{"key":"B1","first-page":"6713","article-title":"\u201cTowards open-set identity preserving face synthesism,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Bao","year":"2018"},{"key":"B2","first-page":"531","article-title":"\u201cMutual information neural estimation,\u201d","volume-title":"Proceedings of the International Conference on Machine Learning","author":"Belghazi","year":"2018"},{"key":"B3","first-page":"172","article-title":"\u201cStochastic gradient descent-ascent: unified theory and new efficient methods,\u201d","volume-title":"Proceedings of the International Conference on Artificial Intelligence and Statistics 2023, Vol. 206","author":"Beznosikov","year":"2022"},{"key":"B4","first-page":"5939","article-title":"\u201cDiffusion facial forgery detection,\u201d","volume-title":"Proceedings of the ACM International Conference on Multimedia","author":"Cheng","year":"2024"},{"key":"B5","first-page":"8789","article-title":"\u201cStargan: unified generative adversarial networks for multi-domain image-to-image translation,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Choi","year":"2018"},{"key":"B6","first-page":"5781","article-title":"\u201cOn the detection of digital face manipulation,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Dang","year":"2020"},{"key":"B7","first-page":"18","article-title":"\u201cExplaining deepfake detection by analysing image matching,\u201d","volume-title":"European Conference on Computer Vision","author":"Dong","year":"2022"},{"key":"B8","first-page":"1180","article-title":"\u201cUnsupervised domain adaptation by backpropagation,\u201d","volume-title":"Proceedings of the International Conference on Machine Learning, Vol. 37","author":"Ganin","year":"2015"},{"key":"B9","first-page":"9821","article-title":"\u201c3D guided fine-grained face manipulation,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Geng","year":"2019"},{"key":"B10","doi-asserted-by":"publisher","first-page":"1263","DOI":"10.1109\/TKDE.2008.239","article-title":"Learning from imbalanced data","volume":"21","author":"He","year":"2009","journal-title":"IEEE Trans. Knowl. Data Eng"},{"key":"B11","article-title":"\u201cLearning deep representations by mutual information estimation and maximization,\u201d","author":"Hjelm","year":"2019","journal-title":"International Conference on Learning Representations"},{"key":"B12","first-page":"3577","article-title":"\u201cImproving the generalization ability of deepfake detection via disentangled representation learning,\u201d","volume-title":"IEEE International Conference on Image Processing","author":"Hu","year":"2021"},{"key":"B13","first-page":"6166","article-title":"\u201cAnti-compression contrastive facial forgery detection,\u201d","volume-title":"IEEE Transactions on Multimedia, Vol. 26","author":"Huang","year":"2023"},{"key":"B14","doi-asserted-by":"crossref","first-page":"720","DOI":"10.1007\/978-3-642-04898-2_327","article-title":"\u201cKullback-leibler divergence,\u201d","volume-title":"International Encyclopedia of Statistical Science","author":"Joyce","year":"2011"},{"key":"B15","article-title":"\u201cProgressive growing of gans for improved quality, stability, and variation,\u201d","volume-title":"International Conference on Learning Representations","author":"Karras","year":"2018"},{"key":"B16","first-page":"4401","article-title":"\u201cA style-based generator architecture for generative adversarial networks,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Karras","year":"2019"},{"key":"B17","first-page":"2828","article-title":"\u201cGeneralized facial manipulation detection with edge region feature extraction,\u201d","volume-title":"Proceedings of the IEEE Winter Conference on Applications of Computer Vision","author":"Kim","year":"2022"},{"key":"B18","doi-asserted-by":"publisher","first-page":"3354","DOI":"10.1073\/pnas.1309933111","article-title":"Equitability, mutual information, and the maximal information coefficient","volume":"111","author":"Kinney","year":"2014","journal-title":"Proc. Nat. Acad. Sci"},{"key":"B19","first-page":"6458","article-title":"\u201cFrequency-aware discriminative feature learning supervised by single-center loss for face forgery detection,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Li","year":"2021"},{"key":"B20","first-page":"5001","article-title":"\u201cFace x-ray for more general face forgery detection,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Li","year":"2020"},{"key":"B21","first-page":"1","article-title":"\u201cIn ICTU oculi: exposing AI created fake videos by detecting eye blinking,\u201d","volume-title":"IEEE International Workshop on Information Forensics and Security","author":"Li","year":"2018"},{"key":"B22","article-title":"\u201cExposing deepfake videos by detecting face warping artifacts,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops","author":"Li","year":"2019"},{"key":"B23","first-page":"3207","article-title":"\u201cCeleb-DF: a large-scale challenging dataset for deepfake forensics,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Li","year":"2020"},{"key":"B24","first-page":"128","article-title":"\u201cExploring disentangled content information for face forgery detection,\u201d","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Liang","year":"2022"},{"key":"B25","first-page":"667","article-title":"\u201cTwo-branch recurrent network for isolating deepfakes in videos,\u201d","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Masi","year":"2020"},{"key":"B26","doi-asserted-by":"publisher","first-page":"307","DOI":"10.1016\/S0016-0032(96)00063-4","article-title":"The jensen-shannon divergence","volume":"334","author":"Men\u00e9ndez","year":"1997","journal-title":"J. Franklin Inst"},{"key":"B27","first-page":"91","article-title":"\u201cOn improving cross-dataset generalization of deepfake detectors,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Nadimpalli","year":"2022"},{"key":"B28","doi-asserted-by":"publisher","first-page":"1038","DOI":"10.1109\/JSTSP.2020.3007250","article-title":"Ganprintr: Improved fakes and evaluation of the state of the art in face manipulation detection","volume":"14","author":"Neves","year":"2020","journal-title":"IEEE J. Sel. Top. Signal Process"},{"key":"B29","first-page":"1","article-title":"\u201cMulti-task learning for detecting and segmenting manipulated facial images and videos,\u201d","volume-title":"IEEE International Conference on Biometrics Theory, Applications and Systems","author":"Nguyen","year":"2019"},{"key":"B30","doi-asserted-by":"publisher","first-page":"103525","DOI":"10.1016\/j.cviu.2022.103525","article-title":"Deep learning for deepfakes creation and detection: a survey","volume":"223","author":"Nguyen","year":"2022","journal-title":"Comput. Vis. Image Underst"},{"key":"B31","doi-asserted-by":"publisher","first-page":"108832","DOI":"10.1016\/j.patcog.2022.108832","article-title":"Learning a deep dual-level network for robust deepfake detection","volume":"130","author":"Pu","year":"2022","journal-title":"Pattern Recognit"},{"key":"B32","first-page":"86","article-title":"\u201cThinking in frequency: face forgery detection by mining frequency-aware clues,\u201d","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Qian","year":"2020"},{"key":"B33","first-page":"1","article-title":"\u201cFaceforensics++: learning to detect manipulated facial images,\u201d","volume-title":"Proceedings of the IEEE International Conference on Computer Vision","author":"Rossler","year":"2019"},{"key":"B34","first-page":"18720","article-title":"\u201cDetecting deepfakes with self-blended images,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Shiohara","year":"2022"},{"key":"B35","first-page":"10096","article-title":"\u201cEfficientnetv2: smaller models and faster training,\u201d","volume-title":"Proceedings of the International Conference on Machine Learning, Vol. 139","author":"Tan","year":"2021"},{"key":"B36","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3306346.3323035","article-title":"Deferred neural rendering: Image synthesis using neural textures","volume":"38","author":"Thies","year":"2019","journal-title":"ACM Trans. Graph"},{"key":"B37","doi-asserted-by":"publisher","DOI":"10.1145\/2929464.2929475","article-title":"\u201cFace2face: real-time face capture and reenactment of rgb videos,\u201d","author":"Thies","year":"2016","journal-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition"},{"key":"B38","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.141","article-title":"\u201cDisentangled representation learning gan for pose-invariant face recognition,\u201d","author":"Tran","year":"2017","journal-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition"},{"key":"B39","first-page":"3527","article-title":"\u201cDecorrelated adversarial learning for age-invariant face recognition,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Wang","year":"2019"},{"key":"B40","doi-asserted-by":"publisher","first-page":"14548","DOI":"10.1609\/aaai.v37i12.26701","article-title":"Noise based deepfake detection via multi-head relative-interaction","volume":"37","author":"Wang","year":"2023","journal-title":"Proc. AAAI Conf. Artif. Intell"},{"key":"B41","doi-asserted-by":"publisher","first-page":"22445","DOI":"10.1109\/ICCV51070.2023.02051","article-title":"Dire for diffusion-generated image detection","author":"Wang","year":"2023","journal-title":"In Proceedings of the IEEE International Conference on Computer Vision"},{"key":"B42","doi-asserted-by":"publisher","first-page":"9005","DOI":"10.1609\/aaai.v33i01.33019005","article-title":"Disentangled variational representation for heterogeneous face recognition","volume":"33","author":"Wu","year":"2019","journal-title":"Proc. AAAI Conf. Artif. Intell"},{"key":"B43","first-page":"848","article-title":"\u201cOptimizing classifier performance via an approximation to the wilcoxon-mann-whitney statistic,\u201d","volume-title":"Proceedings of the International Conference on Machine Learning","author":"Yan","year":"2003"},{"key":"B44","first-page":"22412","article-title":"\u201cUCF: uncovering common features for generalizable deepfake detection,\u201d","volume-title":"Proceedings of the IEEE International Conference on Computer Vision","author":"Yan","year":"2023"},{"key":"B45","doi-asserted-by":"crossref","first-page":"4137","DOI":"10.1145\/3474085.3475546","article-title":"\u201cHeterogeneous face recognition with attention-guided feature disentangling,\u201d","volume-title":"Proceedings of the 29th ACM International Conference on Multimedia","author":"Yang","year":"2021"},{"key":"B46","first-page":"8261","article-title":"\u201cExposing deep fakes using inconsistent head poses,\u201d","volume-title":"IEEE International Conference on Acoustics, Speech and Signal Processing","author":"Yang","year":"2019"},{"key":"B47","doi-asserted-by":"publisher","first-page":"6785","DOI":"10.1109\/TMM.2024.3355651","article-title":"Improving deepfake detection generalization by invariant risk minimization","volume":"26","author":"Yin","year":"2024","journal-title":"IEEE Trans. Multimedia"},{"key":"B48","doi-asserted-by":"publisher","first-page":"127192","DOI":"10.1016\/j.neucom.2023.127192","article-title":"FDML: feature disentangling and multi-view learning for face forgery detection","volume":"572","author":"Yu","year":"2024","journal-title":"Neurocomputing"},{"key":"B49","doi-asserted-by":"publisher","first-page":"547","DOI":"10.1109\/TIFS.2022.3146781","article-title":"Improving generalization by commonality learning in face forgery detection","volume":"17","author":"Yu","year":"2022","journal-title":"IEEE Trans. Inf. Forensics Security"},{"key":"B50","doi-asserted-by":"publisher","first-page":"3405","DOI":"10.1109\/TMM.2023.3310341","article-title":"Narrowing domain gaps with bridging samples for generalized face forgery detection","volume":"26","author":"Yu","year":"2023","journal-title":"IEEE Trans. Multimed"},{"key":"B51","doi-asserted-by":"publisher","first-page":"1499","DOI":"10.1109\/LSP.2016.2603342","article-title":"Joint face detection and alignment using multitask cascaded convolutional networks","volume":"23","author":"Zhang","year":"2016","journal-title":"IEEE Signal Process. Lett"},{"key":"B52","first-page":"2185","article-title":"\u201cMulti-attentional deepfake detection,\u201d","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Zhao","year":"2021"},{"key":"B53","first-page":"15023","article-title":"\u201cLearning self-consistency for deepfake detection,\u201d","volume-title":"Proceedings of the IEEE International Conference on Computer Vision","author":"Zhao","year":"2021"},{"key":"B54","doi-asserted-by":"publisher","first-page":"110077","DOI":"10.1016\/j.patcog.2023.110077","article-title":"Deepfake detection via inter-frame inconsistency recomposition and enhancement","volume":"147","author":"Zhu","year":"2024","journal-title":"Pattern Recognit"}],"container-title":["Frontiers in Big Data"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdata.2025.1669488\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,19]],"date-time":"2025-11-19T21:34:46Z","timestamp":1763588086000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdata.2025.1669488\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,18]]},"references-count":54,"alternative-id":["10.3389\/fdata.2025.1669488"],"URL":"https:\/\/doi.org\/10.3389\/fdata.2025.1669488","relation":{},"ISSN":["2624-909X"],"issn-type":[{"value":"2624-909X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,11,18]]},"article-number":"1669488"}}