Swapped face detection using deep learning and subjective assessment
Open accessAbstract
The tremendous success of deep learning for imaging applications has resulted in numerous beneficial advances. Unfortunately, this success has also been a catalyst for malicious uses such as photo-realistic face swapping of parties without consent. In this study, we use deep transfer learning for face swapping detection, showing true positive rates greater than 96% with very few false alarms. Distinguished from existing methods that only provide detection accuracy, we also provide uncertainty for each prediction, which is critical for trust in the deployment of such detection systems. Moreover, we provide a comparison to human subjects. To capture human recognition performance, we build a website to collect pairwise comparisons of images from human subjects. Based on these comparisons, we infer a consensus ranking from the image perceived as most real to the image perceived as most fake. Overall, the results show the effectiveness of our method. As part of this study, we create a novel dataset that is, to the best of our knowledge, the largest swapped face dataset created using still images. This dataset will be available for academic research use per request. Our goal of this study is to inspire more research in the field of image forensics through the creation of a dataset and initial analysis.
Integrity signals
Registry
Cites
Statistics
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?
Ding et al. (2020)'s swapped face dataset contains 420,053 images of celebrities, including 156,930 real and 263,123 fake face-swapped images created using Nirkin's method and Auto-Encoder-GAN.
“Ding et al. (2020)'s swapped face dataset: This dataset contains 420,053 images of celebrities, including 156,930 real ones downloaded using Google Image API and 263,123 fake face-swapped ones created using two different methods (Nirkin's method and Auto-Encoder-GAN).”
From Deepfake: definitions, performance metrics and standards, datasets, and a meta-review (2024) · cites this paper as Ding et al., 2020
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 · see citing works
SURF: Speeded Up Robust Features
Bay, H., Tuytelaars, T., & Van Gool, L. (2006) · Lecture Notes in Computer Science · doi:10.1007/11744023_32 · Cited by 14,955 · cited 1× · see citing works
“They used sped up robust features (SURF) [4] and Bag of Words (BoW) to create image features instead of using raw pixels.”
MethodShin, H.-C., Roth, H. R., Gao, M., Lu, L., Xu, Z., Nogues, I., Yao, J., Mollura, D., & Summers, R. M. (2016) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2016.2528162 · Cited by 5,759 · see citing works
Wang, G., Li, W., Aertsen, M., Deprest, J., Ourselin, S., & Vercauteren, T. (2019) · Neurocomputing · doi:10.1016/j.neucom.2019.01.103 · Cited by 659 · see citing works
DeepFakes: a New Threat to Face Recognition? Assessment and Detection
Korshunov, P., & Marcel, S. (2018) · arXiv · doi:10.48550/arxiv.1812.08685 · Cited by 497 · see citing works
Face swapping: automatically replacing faces in photographs
Bitouk, D., Kumar, N., Dhillon, S., Belhumeur, P., & Nayar, S. K. (2008) · ACM Transactions on Graphics · doi:10.1145/1360612.1360638 · Cited by 333 · see citing works
Automated face swapping and its detection
Zhang, Y., Zheng, L., & Thing, V. L. L. (2017) · 2017 IEEE 2nd International Conference on Signal and Image Processing (ICSIP) · doi:10.1109/siprocess.2017.8124497 · Cited by 154 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
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 · see citing works
Deepfake Detection: A Systematic Literature Review
Rana, M. S., Nobi, M. N., Murali, B., & Sung, A. H. (2022) · IEEE Access · doi:10.1109/access.2022.3154404 · Cited by 573 · see citing works
Implicit identity leakage: The stumbling block to improving deepfake detection generalization
Dong, S., Wang, J., Ji, R., Liang, J., Fan, H., & Ge, Z. (2023) · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52729.2023.00389 · Cited by 220 · see citing works
Countering Malicious DeepFakes: Survey, Battleground, and Horizon
Juefei-Xu, F., Wang, R., Huang, Y., Guo, Q., Ma, L., & Liu, Y. (2022) · International Journal of Computer Vision · doi:10.1007/s11263-022-01606-8 · Cited by 191 · see citing works
“…of deep learning, various DNN-based approaches are continuously proposed for distinguishing synthesized fake faces, such as deep transfer learning (Ding et al., 2020; Dogonadze et al., 2020; Jeon et al., 2020a), customized CNN (Dang et al., 2019; Marra et al., 2020; Coz-zolino et al., 2018), CNN…”
BackgroundDeepfake forensics analysis: An explainable hierarchical ensemble of weakly supervised models
Silva, S. H., Bethany, M., Votto, A. M., Scarff, I. H., Beebe, N., & Najafirad, P. (2022) · Forensic Science International: Synergy · doi:10.1016/j.fsisyn.2022.100217 · Cited by 66 · see citing works
Deepfake: definitions, performance metrics and standards, datasets, and a meta-review
Altuncu, E., Franqueira, V. N. L., & Li, S. (2024) · Frontiers in Big Data · doi:10.3389/fdata.2024.1400024 · Cited by 65 · see citing works
“’s swapped face dataset [17]: This dataset contains 420,053 images of celebrities, including 156,930 real ones downloaded using Google Image API and 263,123 fake face-swapped ones created using two different methods (Nirkin’s method and Auto-Encoder-GAN) iFakeFaceDB [48]: This dataset includes 87,000 224x224 face imag…”
MethodSupervised Contrastive Learning for Generalizable and Explainable DeepFakes Detection
Xu, Y., Raja, K., & Pedersen, M. (2022) · 2022 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw54805.2022.00044 · Cited by 62 · see citing works
- Authors
- Ding, X., Raziei, Z., Larson, E. C., Olinick, E. V., Krueger, P., & Hahsler, M.
- Affiliations
- Southern Methodist University (US)
- Corresponding author
- Ding, Xinyi
- Venue
- EURASIP Journal on Information Security
- Published
- 2020-05-19
- Volume
- 2020
- Issue
- 1
- ISSN
- 1687-4161 (ISSN-L)
- 2510-523X
- OpenAlex
- W3031973383
- Cited by
- 71
- References
- 7
- Citations by year
- 2026: 7 · 2025: 8 · 2024: 16 · 2023: 8 · 2022: 16 · 2021: 10 · 2020: 6
- FWCI
- 3.89
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- EURASIP Journal on Information Security (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
- Physical Sciences › Computer Science › Signal Processing › Biometric Identification and Security
- Keywords
- deep transfer learning, image forensics, uncertainty estimation, consensus ranking, deepfakes
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
