Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Fair-FLIP: Fair Deepfake Detection with Fairness-Oriented Final Layer Input Prioritising

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Fair-FLIP is a post-processing method that reweights a trained deepfake detector's final-layer activations to reduce ethnic-subgroup disparities, improving fairness metrics by up to 30% while losing only 0.25% accuracy.

desk verdict A genuinely simple post-hoc debiasing idea that might work, but the evaluation protocol needs a clean separation of variance estimation and α tuning from test folds before its headline claim can be trusted. read the letter →

arxiv 2507.08912 v1 pith:3K4DHDCK submitted 2025-07-11 cs.LG cs.AI

Tomasz Szandala , Fatima Ezzeddine , Natalia Rusin , Silvia Giordano , Omran Ayoub This is my paper
classification cs.LGcs.AI
keywords deepfakedetectionfairnessbiasmitigationpost-processingvariancereweightingdemographicparityvisiontransformerprotectedattribute
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a deepfake detector's demographic bias can be reduced in a post-processing step, without retraining and without needing ethnicity labels at inference, by reweighting the features that feed the final classification layer. The method, Fair-FLIP, computes how much each penultimate-layer activation varies across ethnic groups, then demotes the highly variable ones and promotes the stable ones. On a 190,000-image face dataset with a Vision Transformer baseline, the authors report fairness gains of up to 30% in parity metrics—most notably false-positive parity rising from 0.517 to 0.674—while keeping accuracy essentially unchanged at 0.8708 versus 0.8731 for the unmodified model. If these results hold, bias mitigation becomes a lightweight, privacy-preserving add-on for existing detectors rather than a costly retraining exercise.

What carries the argument

The load-bearing object is the normalized between-group standard deviation of each final-layer input, $\hat\sigma(f_i)$, computed from the per-ethnicity mean activations. Fair-FLIP rescales each final-layer weight by $w'_i = w_i \times (1 + \alpha - \hat\sigma(f_i))$, with $\alpha = 0.25$ chosen by sensitivity analysis, so that high-variance features are suppressed and low-variance features are amplified. Because only the final-layer weights are touched and the reweighting is a deterministic function of the activations, the method requires no retraining, no architecture changes, and no protected attributes at inference; its cost is one extra forward pass to collect activations plus a per-feature rescaling.

What would settle it

Apply Fair-FLIP to a detector trained on a dataset where one ethnic group's fake images carry a distinct, high-variance artifact—such as a specific blending trace—and check whether the penalty on that artifact lowers accuracy or fails to improve parity. A more direct test is to shuffle the per-feature penalties while preserving their distribution: if the fairness gain persists under shuffling, the variance ordering itself is not the causal mechanism.

Watch

Extended reading notes

Core claim

The central claim is that final-layer input reweighting based on between-group variability of activations is enough to de-bias a deepfake detector. Specifically, the paper argues that activations whose group-specific means have high standard deviation are biased toward ethnicity-specific features, and that down-weighting them while up-weighting low-variance features improves parity across ethnic groups. The reported experiments, across five-fold cross-validation, show the reweighted model matches baseline accuracy (0.8708 vs 0.8731, a 0.25% drop) while improving all four fairness parities, with the largest effect on false-positive parity (0.5171 to 0.6737). This establishes a new point in the design space: a minimal, inference-time-friendly intervention that improves fairness without the accuracy penalty typical of pruning or threshold methods.

Load-bearing premise

The fairness gain rests on the assumption that a feature's variance across ethnic groups measures how much it encodes ethnicity-specific, bias-prone information; if high-variance features are simply task-relevant cues that happen to differ by group, the reweighting will not improve fairness and may damage detection.

Editorial extensions

If this is right

  • Fairness metrics improve by up to 30% relative, with the largest gain in false-positive parity (0.5171 to 0.6737).
  • Accuracy loss is negligible (0.8708 vs 0.8731 baseline), unlike pre-processing and pruning approaches which lose 1.5–11 points.
  • The method is applicable to any already-trained neural network detector, requires no demographic labels at inference, and adds only a small computational overhead.
  • The authors report that Fair-FLIP preserves the model's attention patterns more closely than BPFA, reducing concern that de-biasing undermines explainability.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the intervention is confined to the final layer, bias encoded in deeper feature hierarchies would likely survive; a layer-wise extension might be needed for models with bias concentrated in mid-level features, at the cost of the method's simplicity.
  • The reported 30% is relative to a baseline with very poor FPP parity, so absolute gains on already-fair models would be smaller; the method's value is greatest where FPP disparities are severe.
  • The same variance-reweighting principle could transfer to other binary classifiers with protected attributes, such as medical screening or fraud detection, if the variance-fairness hypothesis holds there—a testable extension the paper does not perform.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes Fair-FLIP, a post-processing method that mitigates demographic bias in deepfake detection by reweighting the final-layer weights of a trained classifier. The reweighting is based on the between-group standard deviation of penultimate-layer activations across ethnicity groups: features with high between-group variance are demoted, and features with low variance are promoted. Experiments on a Kaggle face dataset with ethnicity labels report that Fair-FLIP improves fairness parity metrics by up to about 30% relative (FPP parity from 0.5171 to 0.6737) while reducing accuracy from 0.8731 to 0.8708, and it compares favorably against pre-processing, in-processing, threshold-based post-processing, and BPFA baselines.

Significance. If the reported results hold, Fair-FLIP is a lightweight, architecture-agnostic fairness intervention that requires no protected attributes at inference time and no retraining, which would be a practically useful contribution to fair deepfake detection. The paper provides open-source code, per-fold results over five-fold cross-validation, and comparisons with several fairness baselines. However, the empirical evidence as presented is not yet sufficient to support the central claim: the method's two data-dependent components, the variance estimates and the hyperparameter α, are not clearly separated from the evaluation folds, so the reported fairness gains may be in-sample rather than generalizable. The contribution is promising but needs a clean train/validation/test protocol and statistical validation before the headline numbers can be accepted.

major comments (3)
  1. [§IV-B, Step 3 and §V-B] The variance estimate σ(f_i) in Step 3 is described as computed from 'each image in the ethnicity-annotated dataset', with no statement that this computation is restricted to the training folds of the five-fold cross-validation. Relatedly, the α sensitivity analysis in §V-B (Fig. 2) reports accuracy and fairness on what appears to be the full dataset, and α=0.25 is selected 'in our opinion' rather than on a held-out validation set. If both the variance statistics and α are derived from the same folds on which the parity metrics in Table I are then reported, Fair-FLIP is effectively tuned on the test distribution, and the headline FPP parity improvement (0.5171→0.6737) could be an in-sample artifact. Please clarify the exact split used for variance estimation and α selection, and ideally re-run the evaluation with a clean train/validation/test separation or nested cross-validation, reporting held-out fairness metrics.
  2. [§V-C, Table I] The paper reports average parity values across five folds but provides no significance tests, confidence intervals, or paired comparisons between Fair-FLIP and the baseline. The per-fold FPP parity improvements are not uniform (e.g., Fold 4: 0.5966→0.6702; Fold 5: 0.4256→0.5989), and the reported standard deviations overlap with baseline in several cases. Without a paired significance test (e.g., a paired t-test or Wilcoxon signed-rank test over folds) or an error analysis on the parity estimators, the claim of a '30% improvement' is not established beyond sampling variation.
  3. [§IV-B] The method's rationale rests on the hypothesis that activations with high between-group variance are biased to ethnicity-specific features, whereas those with lower variance capture more general, non-ethnicity-dependent characteristics. The manuscript provides no independent evidence for this link: no analysis of which features are actually reweighted, no control experiment with randomly reweighted features, and no second dataset. If this hypothesis is false, the reweighting could in principle fail to improve fairness or even harm it. I recommend adding an ablation (e.g., comparing Fair-FLIP against random or inverse-variance reweighting) or additional datasets to support the proposed mechanism.
minor comments (6)
  1. [§IV-B, Step 3] The expression 'σi(fi) = std( ¯fi))' has a mismatched parenthesis and uses σ_i without defining it; please rewrite as σ(f_i) = std(\bar{f}_i) and align Step 4's min/max notation accordingly.
  2. [§IV-B, Step 4] The normalization formula is written as 'σi−min(σ)/max(σ)−min(σ)', which is ambiguous about whether min and max are taken over all features; please clarify the indexing and add parentheses.
  3. [§V-B, Fig. 2] Figure 2 has no axis labels or legend and does not state the full search range of α; please add these details so the sensitivity analysis is reproducible.
  4. [§V-C] The text says the threshold-based approach incurs 'only 4% decline in predictive performance', but Table I shows accuracy 0.8322 versus 0.8731 for the baseline, which is about a 4.7% relative drop; please reconcile or state which quantity is being compared.
  5. [§VI] The explainability analysis is qualitative and appears to be based on a single example per class; please state the number of images inspected and consider a quantitative similarity measure between the attention maps of Baseline, BPFA, and Fair-FLIP.
  6. [Conclusion] The conclusion states 'enhancing fairness metrics by up to 30%' without specifying that this is the relative FPP parity improvement; please specify the metric and the baseline value to avoid overgeneralization.

Circularity Check

1 steps flagged · score 5.0 of 10

Headline fairness gains are selected by tuning α on the evaluated metrics; variance statistics are computed on the full annotated dataset as written.

  1. fitted input called prediction [Section IV-B (Steps 1 and 5, Eq. 6); Section V-B]
    "1) Capture penultimate-layer activations: For each image in the ethnicity-annotated dataset, record the activation vector from second-to-last layer of the deepfake detection model. ... To determine the value of α parameter of Fair-FLIP for the subsequent experiments, we performed a sensitivity analysis varying the value of α. Figure 2 reports the accuracy and fairness metrics achieved for values of α ranging between 0.0 and 1.0. ... For our experiments, we select α = 0.25, which achieves the best balance between accuracy and fairness metrics, in our opinion."

    The reported headline numbers (e.g., FPP parity 0.5171→0.6737, a ~30% relative gain) are the same quantities used to select the hyperparameter α and to estimate between-group variance. Section V-B plots 'accuracy and fairness metrics' against α and picks α=0.25 by 'best balance' without describing any train/validation split, so Table I's results are post-selection. Section IV-B Step 1 says activations are recorded for 'each image in the ethnicity-annotated dataset', not a training fold, meaning the Eq. 6 reweighting can use test-fold statistics in the reported five-fold cross-validation. The conclusion that Fair-FLIP 'can enhance fairness metrics by up to 30%' is therefore a fitted/tuned outcome, not an independent out-of-sample prediction.

full rationale

The core Fair-FLIP update (Eq. 6) is not tautological: it reweights final-layer features according to between-group variance, and this heuristic could in principle hurt fairness, so the algorithm's definition is not self-referential. No load-bearing uniqueness theorem or ansatz is imported from the authors' prior work; the variance-fairness hypothesis is stated as the authors' own assumption. However, the evaluation loop is partially circular. The text does not state that the variance statistics in Section IV-B Step 3 are computed per training fold, and Section V-B selects α by inspecting accuracy and fairness metrics without a described held-out validation split. Consequently, the parities reported in Table I are not clearly out-of-sample; the 30% FPP parity improvement and the 0.25% accuracy preservation are at least partly artifacts of tuning on the evaluation metrics. The limitations section acknowledges single-dataset evaluation but does not address this within-dataset information leakage. Overall this is partial circularity in the reported evidence, not in the method's mathematical derivation.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The ledger is small. The only free parameter is alpha. The key axioms are the variance-fairness hypothesis and the sufficiency of final-layer adjustment. No new entities are introduced. The method is empirical and its claims rest on these assumptions.

free parameters (1)
  • alpha = 0.25
    Selected from a sensitivity analysis on the dataset (Section V-B). Controls the strength of feature promotion/demotion in Equation 6. Chosen by the authors to balance accuracy and fairness, not from a separate validation set.
assumptions (3)
  • domain assumption Activations with low between-group variance are ethnicity-independent, while high variance indicates ethnicity-specific features.
    This is the core hypothesis stated in Section IV-B. The entire method depends on this variance-fairness link, but the paper provides no independent verification.
  • domain assumption The final-layer features are sufficient to adjust fairness without retraining earlier layers.
    Section IV-B claims deeper layers encode more abstract features and that de-biasing the final layer is enough. This is a modeling choice that is not derived or tested across multiple architectures.
  • domain assumption The ethnicity annotations from the HyperExtended LightFace model are accurate enough for the variance computation.
    Section V-A states ethnicity labels come from a 97% accurate facial attribute model with manual verification, but no details on the verification process are given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fair-FLIP: Fair Deepfake Detection with Fairness-Oriented Final Layer Input Prioritising." pith.science (2026). https://pith.science/paper/3K4DHDCK

@misc{pith2026250708912,
  author       = {Pith},
  title        = {Pith review of: Fair-FLIP: Fair Deepfake Detection with Fairness-Oriented Final Layer Input Prioritising},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3K4DHDCK}},
  note         = {Machine review of arXiv:2507.08912}
}
read the original abstract

Artificial Intelligence-generated content has become increasingly popular, yet its malicious use, particularly the deepfakes, poses a serious threat to public trust and discourse. While deepfake detection methods achieve high predictive performance, they often exhibit biases across demographic attributes such as ethnicity and gender. In this work, we tackle the challenge of fair deepfake detection, aiming to mitigate these biases while maintaining robust detection capabilities. To this end, we propose a novel post-processing approach, referred to as Fairness-Oriented Final Layer Input Prioritising (Fair-FLIP), that reweights a trained model's final-layer inputs to reduce subgroup disparities, prioritising those with low variability while demoting highly variable ones. Experimental results comparing Fair-FLIP to both the baseline (without fairness-oriented de-biasing) and state-of-the-art approaches show that Fair-FLIP can enhance fairness metrics by up to 30% while maintaining baseline accuracy, with only a negligible reduction of 0.25%. Code is available on Github: https://github.com/szandala/fair-deepfake-detection-toolbox

Figures

Figures reproduced from arXiv: 2507.08912 by the authors.

Figure 1
Figure 1. Schematic representation of Fair-FLIP’s methodology. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Sensitivity analysis of trade-offs between fairness metrics and accuracy [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Attention heatmaps for the baseline, BPFA, and Fair-FLIP. In the first row (genuine image), baseline and Fair-FLIP match closely, whereas BPFA’s [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Toward Calibrated, Fair, and accurate Deepfake Detection

    cs.LG 2026-06 unverdicted novelty 7.0 of 10

    Face-Feature Tuning is a label-free logit remapping method that reduces FPR/TPR gaps across groups in deepfake detection while preserving overall accuracy.

Reference graph

Works this paper leans on

60 extracted references · 50 canonical work pages · cited by 1 Pith paper

  1. [1]

    A survey of ai-generated content (aigc),

    Y . Cao, S. Li, Y . Liu, Z. Yan, Y . Dai, P. Yu, and L. Sun, “A survey of ai-generated content (aigc),” ACM Computing Surveys , vol. 57, no. 5, pp. 1–38, 2025

    work page 2025

  2. [2]

    Deepfake detection: A systematic literature review,

    M. S. Rana, M. N. Nobi, B. Murali, and A. H. Sung, “Deepfake detection: A systematic literature review,” IEEE access , vol. 10, pp. 25 494–25 513, 2022

    work page 2022

  3. [3]

    Explaining deepfake detection by analysing image matching,

    S. Dong, J. Wang, J. Liang, H. Fan, and R. Ji, “Explaining deepfake detection by analysing image matching,” in European conference on computer vision. Springer, 2022, pp. 18–35

    work page 2022

  4. [4]

    Deep fakes: A looming challenge for privacy, democracy, and national security,

    B. Chesney and D. Citron, “Deep fakes: A looming challenge for privacy, democracy, and national security,”Calif. L. Rev., vol. 107, p. 1753, 2019

    work page 2019

  5. [5]

    On the conversational persuasiveness of large language models: A randomized controlled trial,

    F. Salvi, M. H. Ribeiro, R. Gallotti, and R. West, “On the conversational persuasiveness of large language models: A randomized controlled trial,” arXiv preprint arXiv:2403.14380 , 2024

    arXiv 2024

  6. [6]

    Deepfake detection by analyz- ing convolutional traces,

    L. Guarnera, O. Giudice, and S. Battiato, “Deepfake detection by analyz- ing convolutional traces,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops , 2020

    work page 2020

  7. [7]

    Cyberbullying girls with pornographic deepfakes is a form of misogyny,

    A. M. Narvali, J. A. G. Skorburg, and M. J. Goldenberg, “Cyberbullying girls with pornographic deepfakes is a form of misogyny,” https://theconversation.com/cyberbullying-girls-with-pornographic- deepfakes-is-a-form-of-misogyny-217182, accessed: 2024-11-28

    work page 2024

  8. [8]

    Eros and ethos in celebrity deepfake pornography,

    A. Davisson, “Eros and ethos in celebrity deepfake pornography,” in Ethos, Technology, and AI in Contemporary Society . Routledge, 2024, pp. 186–204

    work page 2024

Show all 60 references
  1. [9]

    US disinformation surge rings alarm bells for Swiss direct democracy,

    Sara Ibrahim, “US disinformation surge rings alarm bells for Swiss direct democracy,” https://www.swissinfo.ch/eng/democracy/ us-disinformation-surge-rings-alarm-bells-for-swiss-direct-democracy/ 87613037, online; accessed 10 January 2025

    2025

  2. [10]

    Deepfake detection for human face images and videos: A survey,

    A. Malik, M. Kuribayashi, S. M. Abdullahi, and A. N. Khan, “Deepfake detection for human face images and videos: A survey,” Ieee Access , vol. 10, pp. 18 757–18 775, 2022

    2022

  3. [11]

    A novel deep learning approach for deepfake image detection,

    A. Raza, K. Munir, and M. Almutairi, “A novel deep learning approach for deepfake image detection,” Applied Sciences, vol. 12, p. 9820, 2022

    2022

  4. [12]

    Deepfake detection using svm,

    H. Agarwal, A. Singh, and D. Rajeswari, “Deepfake detection using svm,” in 2021 Second Int. Conference on Electronics and Sustainable Communication Systems (ICESC) . IEEE, 2021

    2021

  5. [13]

    Interpretable-through- prototypes deepfake detection for diffusion models,

    A. Aghasanli, D. Kangin, and P. Angelov, “Interpretable-through- prototypes deepfake detection for diffusion models,” in Proceedings of the IEEE/CVF international conference on computer vision , 2023

    2023

  6. [14]

    Fairness through awareness,

    C. Dwork, M. Hardt, T. Pitassi, O. Reingold, and R. Zemel, “Fairness through awareness,” in Proceedings of the 3rd innovations in theoretical computer science conference , 2012, pp. 214–226

    2012

  7. [15]

    Extending the theory of information poverty to deepfake technology,

    W. Matli, “Extending the theory of information poverty to deepfake technology,” International Journal of Information Management Data Insights, vol. 4, no. 2, p. 100286, 2024

    2024

  8. [16]

    Fairness and machine learning. fairmlbook. org,

    S. Barocas, M. Hardt, and A. Narayanan, “Fairness and machine learning. fairmlbook. org,” 2019

    2019

  9. [17]

    Equality of opportunity in supervised learning,

    M. Hardt, E. Price, and N. Srebro, “Equality of opportunity in supervised learning,” Advances in neural information processing systems , vol. 29, 2016

    2016

  10. [18]

    Fair prediction with disparate impact: A study of bias in recidivism prediction instruments,

    A. Chouldechova, “Fair prediction with disparate impact: A study of bias in recidivism prediction instruments,” Big data , vol. 5, no. 2, pp. 153–163, 2017

    2017

  11. [19]

    Classifying socially sensitive data without discrimination: An analysis of a crime suspect dataset,

    F. Kamiran, A. Karim, S. Verwer, and H. Goudriaan, “Classifying socially sensitive data without discrimination: An analysis of a crime suspect dataset,” in 2012 IEEE 12th International Conference on Data Mining Workshops. IEEE, 2012, pp. 370–377

    2012

  12. [20]

    Learning fair representations,

    R. Zemel, Y . Wu, K. Swersky, T. Pitassi, and C. Dwork, “Learning fair representations,” in International conference on machine learning . PMLR, 2013, pp. 325–333

    2013

  13. [21]

    Fairtest: Discovering unwarranted associa- tions in data-driven applications,

    F. Tramer, V . Atlidakis, R. Geambasu, D. Hsu, J.-P. Hubaux, M. Hum- bert, A. Juels, and H. Lin, “Fairtest: Discovering unwarranted associa- tions in data-driven applications,” in 2017 IEEE European Symposium on Security and Privacy (EuroS&P) . IEEE, 2017, pp. 401–416

    2017

  14. [22]

    Discriminating feature ratio: Introduc- ing metric for uncovering vulnerabilities in deep convolutional neural networks,

    T. Szandała and H. Maciejewski, “Discriminating feature ratio: Introduc- ing metric for uncovering vulnerabilities in deep convolutional neural networks,” Knowledge-Based Systems, vol. 302, p. 112306, 2024

    2024

  15. [23]

    Fairness-aware configuration of machine learning libraries,

    S. Tizpaz-Niari, A. Kumar, G. Tan, and A. Trivedi, “Fairness-aware configuration of machine learning libraries,” in Proceedings of the 44th International Conference on Software Engineering , 2022, pp. 909–920

    2022

  16. [24]

    Fairness constraints: Mechanisms for fair classification,

    M. B. Zafar, I. Valera, M. G. Rogriguez, and K. P. Gummadi, “Fairness constraints: Mechanisms for fair classification,” in Artificial intelligence and statistics. PMLR, 2017, pp. 962–970

    2017

  17. [25]

    Training well-generalizing classifiers for fairness metrics and other data-dependent constraints,

    A. Cotter, M. Gupta, H. Jiang, N. Srebro, K. Sridharan, S. Wang, B. Woodworth, and S. You, “Training well-generalizing classifiers for fairness metrics and other data-dependent constraints,” in International Conference on Machine Learning . PMLR, 2019, pp. 1397–1405

    2019

  18. [26]

    Censoring representations with an adver- sary,

    H. Edwards and A. Storkey, “Censoring representations with an adver- sary,” arXiv preprint arXiv:1511.05897 , 2015

    2015 arXiv

  19. [27]

    Mitigating unwanted biases with adversarial learning,

    B. H. Zhang, B. Lemoine, and M. Mitchell, “Mitigating unwanted biases with adversarial learning,” in Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society , 2018, pp. 335–340

    2018

  20. [28]

    Calibrating predictions to decisions: A novel approach to multi-class calibration,

    S. Zhao, M. Kim, R. Sahoo, T. Ma, and S. Ermon, “Calibrating predictions to decisions: A novel approach to multi-class calibration,” Advances in Neural Information Processing Systems , vol. 34, pp. 22 313–22 324, 2021

    2021

  21. [29]

    Post-processing fairness evaluation of federated models: An unsupervised approach in healthcare,

    I. Siniosoglou, V . Argyriou, P. Sarigiannidis, T. Lagkas, A. Sarigiannidis, S. K. Goudos, and S. Wan, “Post-processing fairness evaluation of federated models: An unsupervised approach in healthcare,” IEEE/ACM Transactions on Computational Biology and Bioinformatics , vol. 20...

    2023

  22. [30]

    Detection of deepfake video manipulation,

    M. Koopman, A. M. Rodriguez, and Z. Geradts, “Detection of deepfake video manipulation,” in The 20th Irish machine vision and image processing conference (IMVIP), 2018, pp. 133–136

    2018

  23. [31]

    Camera identification by grouping images from database, based on shared noise patterns,

    T. Baar, W. van Houten, and Z. Geradts, “Camera identification by grouping images from database, based on shared noise patterns,” arXiv preprint arXiv:1207.2641, 2012

    2012 arXiv

  24. [32]

    The generalization of ‘student’s’problem when several different population varlances are involved,

    B. L. Welch, “The generalization of ‘student’s’problem when several different population varlances are involved,” Biometrika, vol. 34, no. 1-2, pp. 28–35, 1947

    1947

  25. [33]

    Limits of deepfake detection: A robust estimation viewpoint,

    S. Agarwal and L. R. Varshney, “Limits of deepfake detection: A robust estimation viewpoint,” arXiv preprint arXiv:1905.03493 , 2019

    1905 arXiv

  26. [34]

    Detecting and simulating artifacts in gan fake images,

    X. Zhang, S. Karaman, and S.-F. Chang, “Detecting and simulating artifacts in gan fake images,” in 2019 IEEE international workshop on information forensics and security (WIFS) . IEEE, 2019, pp. 1–6

    2019

  27. [35]

    Detection of deep- fakes using visual artifacts and neural network classifier,

    M. Sahla Habeeba, A. Lijiya, and A. M. Chacko, “Detection of deep- fakes using visual artifacts and neural network classifier,” in Innovations in Electrical and Electronic Engineering: Proceedings of ICEEE 2020 . Springer, 2021, pp. 411–422

    2020

  28. [36]

    Exposing deep fakes using inconsistent head poses,

    X. Yang, Y . Li, and S. Lyu, “Exposing deep fakes using inconsistent head poses,” in ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2019

    2019

  29. [37]

    Deep detection for face manipulation,

    D. Feng, X. Lu, and X. Lin, “Deep detection for face manipulation,” in Neural Information Processing: 27th International Conference, ICONIP 2020, Bangkok, Thailand, November 18–22, 2020, Proceedings, Part V

    2020

  30. [38]

    Springer, 2020, pp. 316–323

    2020

  31. [39]

    Mesonet: a compact facial video forgery detection network,

    D. Afchar, V . Nozick, J. Yamagishi, and I. Echizen, “Mesonet: a compact facial video forgery detection network,” in 2018 IEEE international workshop on information forensics and security (WIFS) . IEEE, 2018

    2018

  32. [40]

    Deepfake detection based on the discrepancy between the face and its context,

    Y . Nirkin, L. Wolf, Y . Keller, and T. Hassner, “Deepfake detection based on the discrepancy between the face and its context,” arXiv preprint arXiv:2008.12262, 2020

    2008 arXiv

  33. [41]

    Deeprhythm: Exposing deepfakes with attentional visual heartbeat rhythms,

    H. Qi, Q. Guo, F. Juefei-Xu, X. Xie, L. Ma, W. Feng, Y . Liu, and J. Zhao, “Deeprhythm: Exposing deepfakes with attentional visual heartbeat rhythms,” in Proceedings of the 28th ACM international conference on multimedia, 2020, pp. 4318–4327

    2020

  34. [42]

    Deepfakeson-phys: Deepfakes detection based on heart rate estimation,

    J. Hernandez-Ortega, R. Tolosana, J. Fierrez, and A. Morales, “Deepfakeson-phys: Deepfakes detection based on heart rate estimation,” arXiv preprint arXiv:2010.00400 , 2020

    2010 arXiv

  35. [43]

    Towards generalizable deepfake detection with locality-aware autoencoder,

    M. Du, S. Pentyala, Y . Li, and X. Hu, “Towards generalizable deepfake detection with locality-aware autoencoder,” in Proceedings of the 29th ACM International Conference on Information & Knowledge Manage- ment, 2020, pp. 325–334

    2020

  36. [44]

    Exploiting visual artifacts to expose deepfakes and face manipulations,

    F. Matern, C. Riess, and M. Stamminger, “Exploiting visual artifacts to expose deepfakes and face manipulations,” in 2019 IEEE Winter Applications of Computer Vision Workshops (WACVW) . IEEE, 2019, pp. 83–92

    2019

  37. [45]

    Tensor-based deepfake detection in scaled and compressed images,

    S. Concas, G. Perelli, G. L. Marcialis, and G. Puglisi, “Tensor-based deepfake detection in scaled and compressed images,” in 2022 IEEE International Conference on Image Processing (ICIP) . IEEE, 2022, pp. 3121–3125

    2022

  38. [46]

    Unmasking deception: Empowering deepfake detection with vision transformer network,

    M. A. Arshed, A. Alwadain, R. Faizan Ali, S. Mumtaz, M. Ibrahim, and A. Muneer, “Unmasking deception: Empowering deepfake detection with vision transformer network,” Mathematics, vol. 11, no. 17, p. 3710, 2023

    2023

  39. [47]

    Fairness evaluation in deepfake detection models using metamorphic testing,

    M. Pu, M. Y . Kuan, N. T. Lim, C. Y . Chong, and M. K. Lim, “Fairness evaluation in deepfake detection models using metamorphic testing,” in Proceedings of the 7th international workshop on metamorphic testing , 2022, pp. 7–14

    2022

  40. [48]

    A survey on bias and fairness in machine learning,

    N. Mehrabi, F. Morstatter, N. Saxena, K. Lerman, and A. Galstyan, “A survey on bias and fairness in machine learning,” ACM computing surveys (CSUR), vol. 54, no. 6, pp. 1–35, 2021

    2021

  41. [49]

    An examination of fairness of ai models for deepfake detection,

    L. Trinh and Y . Liu, “An examination of fairness of ai models for deepfake detection,” arXiv preprint arXiv:2105.00558 , 2021

    2021 arXiv

  42. [50]

    A systematic literature review on the effectiveness of deepfake detection techniques,

    L. Stroebel, M. Llewellyn, T. Hartley, T. S. Ip, and M. Ahmed, “A systematic literature review on the effectiveness of deepfake detection techniques,” Journal of Cyber Security Technology , vol. 7, no. 2, pp. 83–113, 2023

    2023

  43. [51]

    Deepfake: Classifiers, fairness, and demo- graphically robust algorithm,

    A. Agarwal and N. Ratha, “Deepfake: Classifiers, fairness, and demo- graphically robust algorithm,” in 2024 IEEE 18th Int. Conference on Automatic Face and Gesture Recognition (FG) . IEEE, 2024, pp. 1–9

    2024

  44. [52]

    Deepfake detection, im- age manipulation detection, fairness, generalization,

    U. Ezeakunne, C. Eze, and X. Liu, “Deepfake detection, im- age manipulation detection, fairness, generalization,” arXiv preprint arXiv:2412.16428, 2024

    2024 arXiv

  45. [53]

    Improving fairness in deepfake detection,

    Y . Ju, S. Hu, S. Jia, G. H. Chen, and S. Lyu, “Improving fairness in deepfake detection,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2024, pp. 4655–4665

    2024

  46. [54]

    Preserving fairness generalization in deepfake detection,

    L. Lin, X. He, Y . Ju, X. Wang, F. Ding, and S. Hu, “Preserving fairness generalization in deepfake detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 16 815–16 825

    2024

  47. [55]

    Ffb: A fair fairness benchmark for in-processing group fairness methods,

    X. Han, J. Chi, Y . Chen, Q. Wang, H. Zhao, N. Zou, and X. Hu, “Ffb: A fair fairness benchmark for in-processing group fairness methods,” arXiv preprint arXiv:2306.09468 , 2023

    2023 arXiv

  48. [56]

    Thinking racial bias in fair forgery detection: Models, datasets and evaluations,

    D. Liu, Z. Wang, C. Peng, N. Wang, R. Hu, and X. Gao, “Thinking racial bias in fair forgery detection: Models, datasets and evaluations,” arXiv preprint arXiv:2407.14367 , 2024

    2024 arXiv

  49. [57]

    Visualizing and understanding convo- lutional networks,

    M. D. Zeiler and R. Fergus, “Visualizing and understanding convo- lutional networks,” in Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part I 13. Springer, 2014, pp. 818–833

    2014

  50. [58]

    M. Karki. [Online]. Available: https://www.kaggle.com/datasets/ manjilkarki/deepfake-and-real-images

  51. [59]

    Hyperextended lightface: A facial attribute analysis framework,

    S. I. Serengil and A. Ozpinar, “Hyperextended lightface: A facial attribute analysis framework,” in 2021 Int. Conference on Engineering and Emerging Technologies (ICEET) . IEEE, 2021, pp. 1–4

    2021

  52. [60]

    Quantifying attention flow in transformers,

    S. Abnar and W. Zuidema, “Quantifying attention flow in transformers,” arXiv preprint arXiv:2005.00928 , 2020

    2005 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.