Swin-Fake: A Consistency Learning Transformer-Based Deepfake Video Detector
Open accessThe DOI matches its Crossref record; 29 of the 29 references it cites were found in a registry or an index.
Abstract
Deepfake has become an emerging technology affecting cyber-security with its illegal applications in recent years. Most deepfake detectors utilize CNN-based models such as the Xception Network to distinguish real or fake media; however, their performance on cross-datasets is not ideal because they suffer from over-fitting in the current stage. Therefore, this paper proposed a spatial consistency learning method to relieve this issue in three aspects. Firstly, we increased the selections of data augmentation methods to 5, which is more than our previous study’s data augmentation methods. Specifically, we captured several equal video frames of one video and randomly selected five different data augmentations to obtain different data views to enrich the input variety. Secondly, we chose Swin Transformer as the feature extractor instead of a CNN-based backbone, which means that our approach did not utilize it for downstream tasks, and could encode these data using an end-to-end Swin Transformer, aiming to learn the correlation between different image patches. Finally, this was combined with consistency learning in our study, and consistency learning was able to determine more data relationships than supervised classification. We explored the consistency of video frames’ features by calculating their cosine distance and applied traditional cross-entropy loss to regulate this classification loss. Extensive in-dataset and cross-dataset experiments demonstrated that Swin-Fake could produce relatively good results on some open-source deepfake datasets, including FaceForensics++, DFDC, Celeb-DF and FaceShifter. By comparing our model with several benchmark models, our approach shows relatively strong robustness in detecting deepfake media.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Other recent approaches, such as MSIDSnet [23] and Swin-Fake [ 24], employ dual-stream and consistencylearning strategies to fuse spatial, frequency and temporal features within Transformer architectures, yielding notable gains in generalization across manipulation types.”
From Cross-dataset video deepfake detection using Transformer and CNN architectures (2026) · cites this paper as [24]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., & Guo, B. (2021) · 2021 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00986 · Cited by 35,534
Xception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.195 · Cited by 20,380
Generative adversarial networks
Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., & Bengio, Y. (2020) · Communications of the ACM · doi:10.1145/3422622 · Cited by 18,384
Decoupled Weight Decay Regularization
Loshchilov, I., & Hutter, F. (2017) · arXiv · doi:10.48550/arxiv.1711.05101 · Cited by 12,522
A Style-Based Generator Architecture for Generative Adversarial Networks
Karras, T., Laine, S., & Aila, T. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00453 · Cited by 11,714
Denoising Diffusion Probabilistic Models
Ho, J., Jain, A., & Abbeel, P. (2020) · arXiv · doi:10.48550/arxiv.2006.11239 · Cited by 8,712
Random Erasing Data Augmentation
Zhong, Z., Zheng, L., Kang, G., Li, S., & Yang, Y. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i07.7000 · Cited by 3,617
Celeb-DF: A Large-Scale Challenging Dataset for DeepFake Forensics
Li, Y., Yang, X., Sun, P., Qi, H., & Lyu, S. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00327 · Cited by 1,846
Face X-Ray for More General Face Forgery Detection
Li, L., Bao, J., Zhang, T., Yang, H., Chen, D., Wen, F., & Guo, B. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00505 · Cited by 1,211
Research works printed but not linked: 1
- Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., and Gelly, S. (May, January 26). An Image is Worth 16 × 16 Words: Transformers for Image Recognition at Scale. Proceedings of the International Conference on Learning Representations, Virtual.
Cited by
LIDeepDet: Deepfake Detection via Image Decomposition and Advanced Lighting Information Analysis
Lai, Z., Li, J., Wang, C., Wu, J., & Jiang, D. (2024) · Electronics · doi:10.3390/electronics13224466 · Cited by 20
Gong, L. Y., & Li, X. J. (2025) · Electronics · doi:10.3390/electronics14102040 · Cited by 13
Self-citationA Novel Face Swapping Detection Scheme Using the Pseudo Zernike Transform Based Robust Watermarking
Lai, Z., Yao, Z., Lai, G., Wang, C., & Feng, R. (2024) · Electronics · doi:10.3390/electronics13244955 · Cited by 11
MSTransBTS—A Novel Integration of Mamba with Swin Transformer for 3D Brain Tumour Segmentation
Ngu, J. Q., Nisar, H., & Tsai, C. (2025) · Mathematics · doi:10.3390/math13071117 · Cited by 10
Cross-dataset video deepfake detection using Transformer and CNN architectures
Petmezas, G. (2026) · Machine Vision and Applications · doi:10.1007/s00138-026-01809-w · Cited by 6
“Other recent approaches, such as MSIDSnet [23] and Swin-Fake [ 24], employ dual-stream and consistency-learning strategies to fuse spatial, frequency and temporal features within Transformer architectures, yielding notable gains in generalization across manipulation types.”
Chen, N., Xu, T., Sun, M., Yao, C., & Yang, D. (2025) · Intelligent Computing · doi:10.34133/icomputing.0143 · Cited by 5
PBNet: Combining Transformer and CNN in Passport Background Texture Printing Image Classification
Xu, J., Jia, D., Lin, Z., Zhou, T., Wu, J., & Tang, L. B. (2024) · Electronics · doi:10.3390/electronics13214160 · Cited by 3
A Systematic Literature Review on How Cybercriminals Are Utilizing Deepfakes to Conduct Cyberattacks
James, E. C., Bongiovanni, I., Glencross, M., & Brace, L. (2026) · Digital Threats Research and Practice · doi:10.1145/3801545 · Cited by 4
(2025) · International journal of intelligent engineering and systems · doi:10.22266/ijies2025.0930.29 · Cited by 1
Swin-BiLSTM with Attention for Face Swap Detection in Deepfake Videos
Lukitania, N. F., & Suryani, V. (2025) · doi:10.1109/icoict66265.2025.11192953 · Cited by 1
“This implementation utilizes a pretrained Swin Trans-former model (swin base patch4 window7 224) that has had its classification layer removed in order to be used as a feature extractor [14].”
Kotwal, J. G., Pandit, R. J., Dhole, T. S., Chetwani, P. S., Inamdar, B. A., & Pawar, S. D. (2025) · International Conference on Robotics, Automation and Artificial Intelligence (RAAI) · doi:10.1109/raai67517.2025.11423383 · Cited by 2
Chen, J. B., Yang, W., Yin, Y., Lv, H., Wei, F., He, J., Luo, Z., & Wu, J. (2026) · International Conference on Multimedia Retrieval · doi:10.1145/3805622.3810626 · Cited by 1
Adewumi, I. O., & Afe, O. F. (2025) · Research Square · doi:10.21203/rs.3.rs-7793774/v1 · Cited by 1
A Hybrid Approach to Machine Learning-Based Deepfake Video Detection
Esti, M. A. (2025) · Osuva (University of Vaasa) · Cited by 2
Cross-Media Fake Content Detection via Independent Deep Learning Classifiers
Kamal, I. N. B. S., Samsudin, A. S. B., & Hassan, R. B. (2026) · International Journal on Perceptive and Cognitive Computing · doi:10.31436/ijpcc.v12i1.651
“Gong et al. (2024) [11] applied Swin Transformer to video frames, introducing consistency-loss mechanisms to strengthen temporal generalization.”
DeepFake Analysis : A Review of Recent Studies, Detection Methods and Trends in Deepfake Detection
Kaur, P., & Sharma, C. (2025) · International Conference on Innovations and Emerging Technologies in AI & Communication Systems (IETACS) · doi:10.1109/ietacs68750.2025.11385353
“The way makes it easier to see detailed facial features, but it’s not good enough for dealing with highly accurate and detailed changes[7].”
DEEPSHIELD: Vision Transformer based Deepfake Image Detection Model
Raj, K. S., George, C., & Abinaya, K. (2026) · International Conference on Image Processing and Capsule Networks (ICIPCN) · doi:10.1109/icipcn67432.2026.11438643
“Above all, the F1-score curve stabilizes close to 0.98 for both training and validation sets in fig[3], therefore showing strong harmonic balance between precision and recall.”
InfluentialA Spatial-Temporal Learning Framework for Robust Deepfake Video Detection
Rama, G. S., Edem, S., & Padhy, I. (2026) · International Conference on Intelligent Sustainable Systems (ICISS) · doi:10.1109/iciss67859.2026.11453538
“Gong et al. proposed the Swin-Fake model to detect deep-fakes [1].”
InfluentialJeswani, U., & Priyadharshini, S. (2026) · International Conference on Innovative Trends in Information Technology (ICITIIT) · doi:10.1109/icitiit68860.2026.11499629
Deepfake Image Detection Using Transfer Learning TechniquesPreprint
PATEL, H. (2026) · Research Square · doi:10.21203/rs.3.rs-9617329/v1
Deepfake Classification Using an Attention-Based Multi-Domain Transformer Model
Sravani, L., Kumar, C. P., Kumar, K. L. P., Venkatesh, A., & Nithin, B. (2026) · International Journal of Innovative Science and Research Technology (IJISRT) · doi:10.38124/ijisrt/26apr2013
Hariharasudhan, S., Risheeth, P. S. L., & Satyanarayana, G. V. V. (2026) · International Conference on Artificial Intelligence & Machine Learning in Communication and Power Systems (AIMLCPS) · doi:10.1109/aimlcps68702.2026.11542454
Optimizing Deepfake Detection Transformer on FaceForensics++: A Hyperparameter Grid Search Study
Rana, P., Kumar, L., Gupta, R., Dhiman, S., Bagga, V., & Gupta, B. (2026) · International Conference Big Data and Machine Learning · doi:10.1109/icbdml68582.2026.11544685
“While CNNs achieved good intra-dataset accuracy, they often struggled to generalize across datasets [7].”
Banupriya, B. K., & Ruby, A. U. (2026) · International Conference on Inventive Research in Computing Applications (ICIRCA) · doi:10.1109/icirca69024.2026.11570463
“In [12], the authors present a deepfake detector using transformers, with consistency learning, that is more robust to recompression and augmentation.”
Transdefend-VC: A Transformer-Driven Deepfake Defense Mechanism for Secure Video Conferencing
Singh, P., Vairagade, R. S., Pawar, P., & Mishra, S. (2026) · Automation and Remote Control · doi:10.1134/s0005117925601587
HDST-CAF: Adaptive CNN Transformer Frequency Fusion for Deepfake Detection Under Visual Degradation
Laad, S. S., & Smitha Shekar, B. (2026) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2026.0170844
“Gong et al. [32] introduced Swin-Fake with 98.4% accuracy on FF++. Soudy et al. [34] achieved 97% accuracy on FF++ with a three-component pipeline fused via majority voting.”
InfluentialCross-Sample Inconsistency-Based Video Forgery Detection Using Graph Convolutional Networks
Wu, M., Shi, C., & Chen, Y. (2026) · Communications in computer and information science · doi:10.1007/978-981-95-9514-3_20
Hussein, H. A., Shaker, K., & Abdullah, S. (2026) · Babylonian Journal of Machine Learning · doi:10.58496/bjml/2026/012
A lightweight multi-scale fourier enhancement network for image forgery detection
Hou, Y., Miao, Z., Wang, J., Lu, X., Zhang, R., & Li, Y. (2026) · Cybersecurity · doi:10.1186/s42400-026-00681-5
Maghsoudi, S., Sangeeta, S., & Kanwal, N. (2026) · Frontiers in Imaging · doi:10.3389/fimag.2026.1935077
Metadata
- Authors
- Gong, L. Y., Li, X. J., & Chong, P. H. J.
- Affiliations
- Auckland University of Technology (NZ)
- Corresponding author
- Gong, Liang Yu
- Li, Xue Jun
- Venue
- Electronics
- Publisher
- MDPI AG
- Published
- 2024-08-01
- Volume
- 13
- Issue
- 15
- Pages
- 3045
- ISSN
- 2079-9292 (ISSN-L)
- OpenAlex
- W4401210768
- Cited by
- 22
- References
- 28
- Citations by year
- 2026: 13 · 2025: 6 · 2024: 3
- FWCI
- 3.07
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Electronics (publishedVersion, cc-by)
- Tuwhera (Auckland University of Technology) (submittedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Generative Adversarial Networks and Image Synthesis
- Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
- Keywords
- deepfake detection, video forgery detection, consistency learning, Swin Transformer, spatial consistency, data augmentation, cross-dataset generalization, overfitting, feature consistency, cosine distance, transformer-based feature extraction
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.