CoChat

Masked Relation Learning for DeepFake Detection

Yang, Ziming, Liang, Jian, Xu, Yuting, Zhang, Xiao-Yu, He, Ran · 2023 · IEEE Transactions on Information Forensics and Security
Cited by 143

The DOI matches its Crossref record; 40 of the 40 references it cites were found in a registry or an index.

Abstract

DeepFake detection aims to differentiate falsified faces from real ones. Most approaches formulate it as a binary classification problem by solely mining the local artifacts and inconsistencies of face forgery, which neglect the relation across local regions. Although several recent works explore local relation learning for DeepFake detection, they overlook the propagation of relational information and lead to limited performance gains. To address these issues, this paper provides a new perspective by formulating DeepFake detection as a graph classification problem, in which each facial region corresponds to a vertex. But relational information with large redundancy hinders the expressiveness of graphs. Inspired by the success of masked modeling, we propose Masked Relation Learning which decreases the redundancy to learn informative relational features. Specifically, a spatiotemporal attention module is exploited to learn the attention features of multiple facial regions. A relation learning module masks partial correlations between regions to reduce redundancy and then propagates the relational information across regions to capture the irregularity from a global view of the graph. We empirically discover that a moderate masking rate (e.g., 50%) brings the best performance gain. Experiments verify the effectiveness of Masked Relation Learning and demonstrate that our approach outperforms the state of the art by 2% AUC on the cross-dataset DeepFake video detection. Code will be available athttps://github.com/zimyang/MaskRelation.

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.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/tifs.2023.3249566

Cites

40 of 40 references verified0 citation errors · 0 not checkable

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.

A few free checks a day · papers already verified open instantly · results are public

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

Show all 40 references

Research works printed but not linked: 55

  1. 10.5555/2969033.2969125 · doi:10.5555/2969033.2969125
  2. Kingma, Auto-encoding variational Bayes, 1
  3. 2016, Fakeapp
  4. 10.1109/ICCV.2019.00728 · DOI · Paper page
  5. Agarwal, Protecting world leaders against deep fakes, 38
  6. Dosovitskiy, An image is worth 16×16 words: Transformers for image recognition at scale, 1
  7. 10.1109/ICASSP43922.2022.9746888 · DOI
  8. 10.1109/CVPR52688.2022.00925 · DOI · Paper page
  9. 10.1609/aaai.v36i1.19954 · DOI · Paper page
  10. 10.1007/978-3-031-20065-6_23 · DOI
  11. 10.1007/978-3-031-19781-9_7 · DOI
  12. 10.1145/3394171.3414034 · DOI · Paper page
  13. 10.1016/j.patcog.2021.107950 · DOI
  14. 10.1016/j.sigpro.2021.108051 · DOI
  15. 10.1016/j.knosys.2022.109114 · DOI
  16. 10.1145/3534678.3539321 · DOI · Paper page
  17. 10.24963/ijcai.2019/565 · DOI
  18. 10.1016/j.patrec.2021.04.021 · DOI
  19. Jin, 2020, Self-supervised learning on graphs: Deep insights and new direction, arXiv:2006.10141 · arXiv
  20. 10.1145/3437880.3460400 · DOI · Paper page
  21. 10.1109/ICCVW.2019.00213 · DOI
  22. 10.1145/3197517.3201350 · DOI
  23. 10.1109/CVPR46437.2021.00361 · DOI · Paper page
  24. 10.1007/978-3-031-19775-8_35 · DOI
  25. 10.1109/TIFS.2020.3013186 · DOI
  26. 10.1609/aaai.v34i07.6857 · DOI
  27. 10.1109/CVPR52688.2022.00304 · DOI
  28. 10.1109/CVPR52688.2022.01515 · DOI
  29. Kipf, Semi-supervised classification with graph convolutional networks, 1
  30. 10.1038/s41467-021-23415-2 · DOI
  31. 10.1016/j.patcog.2022.108581 · DOI
  32. 10.1109/ICASSP.2018.8462291 · DOI · Paper page
  33. 10.1145/1390156.1390160 · DOI
  34. 10.1109/ICCV.2019.00356 · DOI · Paper page
  35. 10.1109/TITS.2019.2935152 · DOI · Paper page
  36. 10.48550/ARXIV.1706.03762 · DOI · Paper page
  37. 10.1109/ICCV.2019.00557 · DOI
  38. Wettig, 2022, Should you mask 15% in masked language modeling?, arXiv:2202.08005 · arXiv
  39. Feichtenhofer, 2022, Masked autoencoders as spatiotemporal learners, arXiv:2205.09113 · arXiv
  40. 10.1109/ICTAI50040.2020.00154 · DOI
  41. 10.1145/3394486.3403140 · DOI · Paper page
  42. 10.1109/CVPR.2018.00675 · DOI · Paper page
  43. 10.1007/978-3-030-58565-5_10 · DOI · Paper page
  44. Krizhevsky, ImageNet classification with deep convolutional neural networks, 1106
  45. Guan, Delving into sequential patches for deepfake detection, 15
  46. 10.1109/CVPR52688.2022.01171 · DOI
  47. 10.1109/ASRU46091.2019.9003738 · DOI
  48. 10.1109/ICCV48922.2021.01211 · DOI
  49. Srivastava, 2014, Dropout: A simple way to prevent neural networks from overfitting, J. Mach. Learn. Res., 15, 1929
  50. 10.1109/CVPR.2016.262 · DOI · Paper page
  51. 10.1145/3306346.3323035 · DOI · Paper page
  52. 10.1109/LSP.2016.2603342 · DOI · Paper page
  53. Kingma, Adam: A method for stochastic optimization, 1
  54. Battaglia, 2018, Relational inductive biases, deep learning, and graph networks, arXiv:1806.01261 · arXiv
  55. 10.1109/TIP.2020.2994412 · DOI

Cited by

See all 143

Metadata

Authors
Yang, Z., Liang, J., Xu, Y., Zhang, X.-Y., & He, R.
Venue
IEEE Transactions on Information Forensics and Security
Published
2023
Volume
18
Pages
1696-1708
ISSN
  • 1556-6013 (ISSN-L)
  • 1556-6021
OpenAlex
W4322576470
Cited by
143
References
40
FWCI
13.00
Type
journal article
Language
en
Open access
closed
Funders
  • National Natural Science Foundation of China (U2003111, U21B2045, 62276256)
  • Beijing Nova Program (Z211100002121108)
  • Chinese Association for Artificial Intelligence (CAAI)-Huawei MindSpore Open Fund

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Masked Relation Learning for DeepFake Detection. https://app.cochat.ai/papers/10.1109/tifs.2023.3249566
The paper
Yang, Z., Liang, J., Xu, Y., Zhang, X.-Y., & He, R. (2023). Masked Relation Learning for DeepFake Detection. IEEE Transactions on Information Forensics and Security. https://doi.org/10.1109/tifs.2023.3249566

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.