OST: Improving Generalization of DeepFake Detection Via One-Shot Test-Time Training
The DOI matches its Crossref record.
Abstract
State-of-the-art deepfake detectors perform well in identifying forgeries when they are evaluated on a test set similar to the training set, but struggle to maintain good performance when the test forgeries exhibit different characteristics from the training images, e.g., forgeries are created by unseen deepfake methods. Such a weak generalization capability hinders the applicability of current deepfake detectors. In this paper, we introduce a new learning paradigm specially designed for the generalizable deepfake detection task. Our key idea is to construct a testsample-specific auxiliary task to update the model before applying it to the sample. Specifically, we synthesize pseudo-training samples from each test image and create a test-time training objective to update the model. Moreover, we propose to leverage meta-learning to ensure that a fast single-step test-time gradient descent, dubbed one-shot test-time training (OST), can be sufficient for good deepfake detection performance. Extensive results across several benchmark datasets demonstrate that our approach performs favorably against existing arts in terms of generalization to unseen data and robustness to different post-processing steps.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“Generalization capability is a critical concern as detectors trained on specific datasets often fail to maintain performance when exposed to unseen data distributions, novel manipulation techniques, or different generator architectures [17-20].”
From Deepfake detection across image, video, and audio: a comprehensive survey with empirical evaluation of generalization and robustness (2026) · cites this paper as [20]
In the citing paper
“Common used techniques include meta-learning [20, 182-184], incremental learning [185], knowledge distillation [186], and domain generalization [187].”
From Deepfake detection across image, video, and audio: a comprehensive survey with empirical evaluation of generalization and robustness (2026) · cites this paper as [20]
In the citing paper
“To mitigate this, OST [20] applies the test-time-training technique [208] that enables detectors to adapt to unknown techniques at inference time.”
From Deepfake detection across image, video, and audio: a comprehensive survey with empirical evaluation of generalization and robustness (2026) · cites this paper as [20]
In the citing paper
“Furthermore, it is important to note that the validation set is not utilized in our experiments. Following previous works [3, 4], we adopt the practice of selecting the model that achieves the highest performance on the test set rather than the validation set for final evaluation.”
From DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection (2023) · cites this paper as [4]
Citations
Cites
Still being checked.
Cited by
Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection
Yan, Z., Luo, Y., Lyu, S., Liu, Q., & Wu, B. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52733.2024.00858 · Cited by 166
DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection
Yan, Z., Zhang, Y., Yuan, X., Lyu, S., & Wu, B. (2023) · arXiv · doi:10.48550/arxiv.2307.01426 · Cited by 97
Self-citationDeepfakeBench: A Comprehensive Benchmark of Deepfake Detection
Yan, Z., Zhang, Y., Yuan, X., Lyu, S., & Wu, B. (2023) · Advances in Neural Information Processing Systems 36 · doi:10.52202/075280-0201 · Cited by 98
Self-citationNguyen-Le, H.-H., Tran, V.-T., Nguyen, D.-T., & Le-Khac, N.-A. (2026) · Artificial Intelligence Review · doi:10.1007/s10462-026-11608-4 · Cited by 28
Vulnerability-Aware Spatio-Temporal Learning for Generalizable Deepfake Video Detection
Nguyen, D., Astrid, M., Kacem, A., Ghorbel, E., & Aouada, D. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.01004 · Cited by 37
UniForensics: Face Forgery Detection via General Facial Representation
Fang, Z., Zhao, H., Wei, T., Zhou, W., Wan, M., Wang, Z., Zhang, W., & Yu, N. (2025) · IEEE Transactions on Dependable and Secure Computing · doi:10.1109/tdsc.2025.3627420 · Cited by 16
LEDNet: a multimodal foundation model for robust deepfake detection
Tao, R., Tang, S., Qin, H., Wang, W., Wei, Y., & Zhao, Y. (2025) · Science China Information Sciences · doi:10.1007/s11432-024-4400-8 · Cited by 9
UniForensics: Face Forgery Detection via General Facial Representation
Fang, Z., Zhao, H., Wei, T., Zhou, W., Wan, M., Wang, Z., Zhang, W., & Yu, N. (2024) · arXiv · doi:10.48550/arxiv.2407.19079 · Cited by 8
Xu, K., Hu, X., Chen, C., Zhou, Y., Xu, Y., Alsolai, H., & Mumtaz, S. (2026) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2026.3690601 · Cited by 4
Associative Recurrent Bilinear Optimization for Domain-Generalized Binary Neural Networks
Xu, S. H., Li, Y., Wang, M., Liu, C., & Zhang, B. (2026) · International Journal of Computer Vision · doi:10.1007/s11263-026-02762-x · Cited by 1
Face Deepfake Detection Based on Up-Sampling Feature Extraction and CNN
Tang, S. (2025) · Communications in computer and information science · doi:10.1007/978-3-031-94962-3_12
Multimodal Semantic-Similarity Gate for Real-Time Optimized Deepfake Detection
Yeon, Y., Yoon, J., & Choi, C. (2026) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2026.3666953 · Cited by 1
Black-Box Adaptation for Deepfake Detection via Local Relation Guided AUC Optimization
Si, X., Li, L., Li, B., Zhang, L., 郭子铎, Yuan, K., & Tian, Q. C. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3686234
Hierarchical Adaptive networks with Task vectors for Test-Time Adaptation
Ambekar, S., Hasny, M., Daza, L. A., Lang, D. M., & Schnabel, J. A. (2026) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61042.2026.00453 · Cited by 3
Evaluating Cross-Dataset Robustness of Video Deepfake Detection Models
Jyothi, B. N., & Jabbar, M. A. (2026) · IEEE International Conference for Convergence in Computing Technology (I3CTCON) · doi:10.1109/i3ctcon68242.2026.11507503
Continual Test-Time Adaptation via Low-Frequency Modulation
Zhang, B., Pan, J., Yang, S., & Gu, L. (2026) · IEEE Access · doi:10.1109/access.2026.3693209
Hota, D., Barik, R. C., Sahoo, S., Padhan, S., & Behera, N. (2026) · International Conference on Artificial Intelligence & Machine Learning in Communication and Power Systems (AIMLCPS) · doi:10.1109/aimlcps68702.2026.11542296
Cover-Source Mismatch in Deepfake Detection: A Systematic Study
Franců, J., Papež, M., & Pevný, T. (2026) · ACM Workshop on Information Hiding and Multimedia Security (IH&MMSec) · doi:10.1145/3785353.3815097
A dual-domain spatio-temporal and frequency framework for robust deepfake detection
Sajjadi, A., Sarvar, S., Nekou, M., Fallah, M., Mehralizadeh, D., Jabbarzadeh, M. H., & Salehpour, P. (2026) · Scientific Reports · doi:10.1038/s41598-026-65079-2
Wtmsa-net: Wavelet transform-based multi-stage attention network for deepfake detection
Chen, Z., Wang, L., Shi, H., Zhang, S., & Liu, H. (2026) · Pattern Analysis and Applications · doi:10.1007/s10044-026-01752-6
FreDA: Training-Free Test-Time Adaptation for Deepfake Detection via Non-parametric Cache Retrieval
Yang, F., Le‐Khac, N., & Nguyen-Le, H.-H. (2026) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-032-37274-1_4
Metadata
- Authors
- Chen, L., Zhang, Y., Song, Y., Wang, J., & Liu, L.
- ORCID
- Chen, Liang
- Venue
- Advances in Neural Information Processing Systems 35
- Published
- 2022
- Pages
- 24597-24610
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W7133198189
- Cited by
- 12
- References
- 0
- Citations by year
- 2026: 10 · 2025: 2
- FWCI
- 0.18
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- neural information processing systems (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Computer Science › Signal Processing › Biometric Identification and Security
- Keywords
- deepfake detection, one-shot learning, test-time training, generalization, video forgery detection
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.