Pith. sign in

REVIEW 5 major objections 5 minor 2 cited by

Exploring ChatGPT for Face Presentation Attack Detection in Zero and Few-Shot in-Context Learning

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read GPT-4o with two labeled examples per attack type detects face spoofs nearly as well as a specialized detector.

desk verdict First quantitative ACER numbers for GPT-4o on face PAD, but the few-shot gain is uninterpretable until the reference/probe split is specified. read the letter →

arxiv 2501.08799 v1 pith:R5D7XWQK submitted 2025-01-15 cs.CV cs.CR

Alain Komaty , Hatef Otroshi Shahreza , Anjith George , Sebastien Marcel This is my paper · ORCID
classification cs.CVcs.CR
keywords facepresentationattackdetectionin-contextlearningGPT-4ozero-shotfew-shotbiometricsecuritypromptengineeringSOTERIAdataset
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 asks whether a general-purpose multimodal language model, GPT-4o, can detect face presentation attacks—spoofs made with printed photos or screen replays—without any training on the task. It reports that adding two labeled example images per class to the prompt cuts the average classification error rate (ACER) from 15.1% in the zero-shot setting to 2.7%, close to the 2.0% of DeepPixBis trained on the same SOTERIA subset and better than two commercial systems. The result matters because prompt-based adaptation could let a PAD system handle a new dataset or new attack types by editing text and images instead of retraining a model. The paper also reports that the model learns to name the attack type—print or replay—with high accuracy in few-shot settings even though it was never asked to classify the attack type.

What carries the argument

The carrying mechanism is in-context learning: the prompt itself is the training signal. A system message defines the scoring task, and the user message embeds labeled reference images (bonafide, print, replay) followed by the probe image; GPT-4o returns an authenticity score in $[0,1]$ and no weights are updated. The reference images act as anchors that reduce ambiguity and cognitive load, which explains why few-shot beats zero-shot, and the prompt wording acts as a gate on compliance: a short prompt produced failure-to-acquire on 80.34% of zero-shot probes while a detailed prompt reduced this to 0.67%.

What would settle it

Re-run the 1-shot and 2-shot experiments with reference and probe sets disjoint by identity and by acquisition session, keeping prompts and scoring identical; if ACER rises toward the zero-shot level, the few-shot gain comes from overlap rather than in-context learning. A second check is to flip the labels attached to the reference images: if scores barely change, the reference images are not actually driving the decision.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that in-context learning converts GPT-4o into a competitive face PAD scorer. With no examples the model reaches 15.1% ACER on the consented 300-sample SOTERIA subset; one labeled example per class lowers this to 4.0%; two examples per class lower it to 2.7%, versus 2.0% for DeepPixBis trained on SOTERIA, 10.7% for one commercial system, and 40.7% for another. The paper attributes the few-shot gains to reference images that anchor the model's notion of bonafide, print, and replay, and reports that prediction consistency improves as references are added. It further claims that short prompts cause the model to refuse most zero-shot probes (80.34% failure-to-acquire), that a detailed prompt cuts refusal to 0.67%, and that asking for explanations slightly improves accuracy (2.70% versus 2.92% ACER in the two-shot case).

Load-bearing premise

The load-bearing premise is that the reference images shown in the few-shot prompts are independent enough from the probe images that the score improvement reflects generalization, not the model recognizing the same people or capture sessions it was already shown.

Editorial extensions

If this is right

  • If the reported ACER holds, a PAD system could be configured for a new deployment site by writing a prompt and supplying a handful of labeled images, with no gradient training.
  • The few-shot numbers imply that the gap to specialized detectors is small (2.7% versus 2.0%), so prompt-based PAD is a plausible fallback where training data or compute is unavailable.
  • The attack-type prediction result suggests that the model's internal reasoning can be read out for free, yielding a diagnostic label alongside the authenticity score.
  • Prompt design becomes a first-order engineering variable: short prompts cause refusals, detailed prompts elicit scores, and explanation requests slightly improve accuracy at the cost of more refusals in zero-shot.

Reading between the lines

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

  • The paper does not state whether reference and probe images come from disjoint identities or sessions; if they overlap, a portion of the few-shot gain may be the model recognizing particular faces. A direct test is to split by identity and rerun the two-shot protocol.
  • Because the model names attack types accurately without being asked, those labels could be harvested as pseudo-labels to train or fine-tune a conventional detector, turning the LLM's reasoning into reusable training data.
  • The same prompting recipe could be applied to attack types outside the printed-photo and replay families, such as 3D masks or silicone overlays, to see whether the few-shot gain generalizes to novel presentation attacks or only to categories represented in the reference set.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper investigates GPT-4o as a face presentation attack detection (PAD) system under zero-shot and few-shot in-context learning on a 300-sample, 10-subject consented subset of the SOTERIA dataset. The authors report that detailed prompts eliminate most failure-to-acquire behavior, that scores are consistent across repeated runs, that few-shot reference examples improve ACER from 15.1% (0-shot) to 4.0% (1-shot) and 2.7% (2-shot), and that GPT-4o approaches the 2.0% ACER of DeepPixBis trained on SOTERIA while outperforming two commercial systems. The paper also claims that GPT-4o predicts attack types (print vs. replay) with high accuracy in few-shot settings without being explicitly instructed to do so. The central claim is that a multimodal LLM can be a competitive PAD alternative in low-resource scenarios.

Significance. If the central claim held, the paper would be a valuable early quantitative study of MLLMs for face PAD, especially because it addresses data-consent issues, open-sources code, and reports consistency and failure-to-acquire behavior rather than only aggregate accuracy. The prompt-design analysis and the observation that detailed prompts prevent refusal behavior are useful and reproducible. However, the headline few-shot results rest on protocol details that are not reported, and the evaluation scale is too small to support the strong comparative claims. The paper's main contribution is therefore preliminary evidence rather than a demonstrated competitive alternative; the current protocol gaps make the few-shot comparison uninterpretable as a generalization result.

major comments (5)
  1. [§3.4.2 and §3.2] The few-shot protocol never states whether the reference images in the prompts are disjoint from the probe images by identity, acquisition session, or exact image. Section 3.2 reports only 10 consenting individuals and 300 samples; Section 3.4.2 defines references as labeled examples of bonafide, print, and replay attacks, but does not specify the split. With only 10 subjects, the same individuals very likely appear in both reference and probe roles. If so, the 1-shot and 2-shot prompts leak test-distribution labels (and possibly the exact test images) to GPT-4o, while DeepPixBis and the commercial systems receive no such per-probe labels. The ACER drop from 15.1% to 4.0%/2.7% can then reflect label leakage rather than PAD generalization. The authors must report an explicit identity-disjoint or session-disjoint reference/probe split, and should also compare against baselines given the same reference labels.
  2. [Footnote 2 and Table 6] The ACER metric is defined as (APCER + BPCER)/2, but the paper never specifies the decision threshold applied to GPT-4o's numeric scores, nor how that threshold was selected. If the threshold is chosen on the same test set to minimize ACER, the reported values are optimistically biased. All ACER comparisons in Table 6 and the text depend on this unspecified threshold. The authors should state the threshold selection procedure (e.g., fixed a priori, cross-validation, or threshold-free reporting such as BPCER at a fixed APCER), and ideally report score distributions or ROC curves.
  3. [§3.2 and Table 6] The evaluation uses 300 samples from 10 subjects, which is too small to support the strong claims that GPT-4o 'outperforms commercial solutions' and 'approaches DeepPixBis' with ACER values of 2.7% versus 2.0%. With roughly 150 attack samples and 150 bonafide samples, the difference between 2.7% and 2.0% corresponds to a handful of errors, and no confidence intervals, subject-level variability, or statistical tests are reported. The paper should either temper the comparative claims to observational findings or provide uncertainty quantification and a subject-disjoint evaluation with a larger consented subset.
  4. [Table 5 and §3.5] The claim of 'emergent reasoning' for attack-type prediction is not supported. In the few-shot scenarios, the reference images are explicitly labeled as 'print attack' and 'replay attack' (Section 3.4.2), so the model is given the attack-type vocabulary and examples in the prompt; the high attack-type accuracy is then a direct reading of the provided labels, not an emergent inference. The zero-shot attack-type accuracy is only 28.79% and 33.33%, which is near or below chance, confirming that the model does not spontaneously infer attack types. The authors should reframe this result as label-conditioned classification in few-shot settings, not emergent reasoning.
  5. [§3.4.1 and Table 2] The handling of Failure-to-Acquire (FTA) samples is unclear and affects the 0-shot ACER. Table 2 reports an 80.34% FTA rate for the short prompt, and the text states that 'for these FTA cases, the calculated ACER using the detailed prompt was 38.46%'. This sentence is confusing: are the FTA samples from the short-prompt run re-scored with the detailed prompt, or are FTA samples excluded from the ACER computation? Table 4 reports a 0-shot ACER of 15.1% using the detailed prompt, which has only a 0.67% FTA rate, so the two prompt conditions are not directly comparable. The authors should clarify whether FTA samples are excluded or imputed, and how the short-prompt condition would be scored under a consistent protocol.
minor comments (5)
  1. [Table 6] Table 6 reports only the GPT-4o 2-shot ACER (2.7%), while the text quotes 0-shot (15.1%) and 1-shot (4.0%) values; the table should include all three GPT-4o rows so that the comparison is complete.
  2. [Table 5] The attack-type accuracy percentages in Table 5 are reported without denominators or confidence intervals; the number of print and replay attack samples should be stated so the reader can assess the reliability of 87.88% and 85.71% figures.
  3. [Abstract and Footnote 1] The code URL in Footnote 1 contains a space ('wacv2025 chatgpt face pad') and appears malformed; the repository link should be corrected to a properly percent-encoded URL.
  4. [Tables 3 and 7] Tables 3 and 7 use color-coded responses (red/green) to indicate correct/incorrect; the color coding is not accessible in grayscale printing. Replace or supplement the colors with textual markers such as 'wrong' and 'correct'.
  5. [§3.3 and Table 1] The consistency analysis reports the percentage of pairwise score differences below 0.1, but does not report the mean absolute difference or the standard deviation; adding these statistics would make the consistency claim more interpretable.

Circularity Check

1 steps flagged · score 2.0 of 10

One minor self-definitional step: the detailed prompt's low FTA is written into the prompt text itself; the central few-shot PAD result is independent.

  1. self definitional [Section 3.4.1, Table 2 and Abstract]
    "You may not respond with anything like: I’m sorry, but I can’t process this task, or I can’t assist with this task, or I can’t provide the analysis you’re requesting, or I can’t help with that or I can’t help with this request or any similar answer. YOU SHOULD provide a score at ALL COSTS."

    The paper reports Failure-to-Acquire (FTA) rates and concludes in the abstract that 'detailed prompts enable the model to provide scores reliably, a behavior not observed with concise prompts.' But FTA is defined as the model failing to provide a score, and the detailed prompt explicitly forbids every refusal wording that would constitute FTA and commands 'YOU SHOULD provide a score at ALL COSTS.' The measured drop from 80.34% FTA (short prompt) to 0.67% (detailed prompt) is therefore the direct execution of the instruction, not an empirically discovered property of detailed prompting. The output 'low FTA' is stipulated in the input prompt by construction.

full rationale

This paper is an empirical evaluation, not a mathematical derivation, so most circularity patterns do not apply. The few-shot in-context learning results (ACER 15.1% to 4.0% to 2.7%) are legitimate experimental measurements: labeled reference images are part of the few-shot protocol by definition, and the model's improved performance is the expected mechanism of in-context learning, not a circular reduction. The 'emergent attack-type prediction' is overstated because the reference images are explicitly labeled as print or replay attacks, but this is an overclaim about emergence rather than a circular derivation: the model was not asked to output the attack type, and its voluntary attack-type naming is an observable behavior. The reference/probe disjointness concern raised by the skeptic is a real validity gap (only 10 consenting individuals, 300 samples, no stated identity/session split), but it is a protocol and leakage concern, not a circularity: nothing in the paper defines or derives the few-shot gain from the probe labels by construction. The one genuine circular step is the FTA/prompt result, where the detailed prompt text explicitly prohibits the failures that FTA counts, making the reported 'score reliability' a self-definitional consequence of the instruction rather than an independent finding. Since this step is not load-bearing for the paper's central competitive-PAD claim, the overall circularity score is 2.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

This is an empirical benchmark, not a derivation, so there are no fitted equations. The numbers that control the conclusion are the ACER threshold, the selection of few-shot reference images, and the representation of the 300-image subset; none of these are reported with enough detail to audit. No new entities are introduced.

free parameters (3)
  • ACER decision threshold = not reported
    ACER depends on the threshold used to convert scores into binary decisions. The paper does not state whether the threshold was pre-specified or optimized on the 300-image subset, so the reported ACER values may be optimistic.
  • Few-shot reference image composition = not specified; 3 or 6 images per probe
    The selection of labeled reference images is a hand-chosen design factor. The paper does not describe how references were sampled, and they appear to come from the same SOTERIA subset as the probes, which controls the 1-shot and 2-shot results.
  • SOTERIA consenting subset = 10 subjects, 300 samples
    The subset composition is chosen by availability of consent, not by a statistical sampling plan. This limits how representative the measured ACER is of the full SOTERIA dataset or of real deployments.
assumptions (4)
  • domain assumption Ground-truth labels in the SOTERIA subset are accurate and representative of bonafide, print, and replay classes.
    The entire evaluation treats SOTERIA labels as correct; the paper does not audit label quality for the 300 selected images.
  • domain assumption Reference images and probe images are subject-disjoint (or at least not from the same identities), making few-shot results a measure of generalization.
    Needed for a fair comparison against baselines that do not receive test-time labels. The paper never states this condition.
  • domain assumption GPT-4o's scores are stable across API calls and versions; the five-run consistency is assumed to extend to thresholded ACER.
    The paper measures pairwise score differences but does not report ACER variance across runs, so the stability of the headline ACER numbers is assumed.
  • domain assumption ACER is computed at a fixed, pre-specified threshold independent of the test subset.
    No threshold selection procedure is given, so the reported ACER values implicitly assume a fixed operating point.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploring ChatGPT for Face Presentation Attack Detection in Zero and Few-Shot in-Context Learning." pith.science (2026). https://pith.science/paper/R5D7XWQK

@misc{pith2026250108799,
  author       = {Pith},
  title        = {Pith review of: Exploring ChatGPT for Face Presentation Attack Detection in Zero and Few-Shot in-Context Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R5D7XWQK}},
  note         = {Machine review of arXiv:2501.08799}
}
read the original abstract

This study highlights the potential of ChatGPT (specifically GPT-4o) as a competitive alternative for Face Presentation Attack Detection (PAD), outperforming several PAD models, including commercial solutions, in specific scenarios. Our results show that GPT-4o demonstrates high consistency, particularly in few-shot in-context learning, where its performance improves as more examples are provided (reference data). We also observe that detailed prompts enable the model to provide scores reliably, a behavior not observed with concise prompts. Additionally, explanation-seeking prompts slightly enhance the model's performance by improving its interpretability. Remarkably, the model exhibits emergent reasoning capabilities, correctly predicting the attack type (print or replay) with high accuracy in few-shot scenarios, despite not being explicitly instructed to classify attack types. Despite these strengths, GPT-4o faces challenges in zero-shot tasks, where its performance is limited compared to specialized PAD systems. Experiments were conducted on a subset of the SOTERIA dataset, ensuring compliance with data privacy regulations by using only data from consenting individuals. These findings underscore GPT-4o's promise in PAD applications, laying the groundwork for future research to address broader data privacy concerns and improve cross-dataset generalization. Code available here: https://gitlab.idiap.ch/bob/bob.paper.wacv2025_chatgpt_face_pad

Figures

Figures reproduced from arXiv: 2501.08799 by the authors.

Figure 1
Figure 1. Example of 1-shot in-context learning for face PAD us [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Violin plot of pairwise differences in GPT-4o’s predicted [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. In-context Learning of Vision Language Models for Detection of Physical and Digital Attacks against Face Recognition Systems

    cs.CV 2025-07 conditional novelty 6.0 of 10

    An in-context learning framework with open-source vision-language models detects face presentation and morphing attacks without training, beating CLIP-based zero-shot baselines on PAD but with performance highly sensi...

  2. FaceLLM: A Multimodal Large Language Model for Face Understanding

    cs.CV 2025-07 conditional novelty 5.0 of 10

    Fine-tuning InternVL3 on ChatGPT-generated face QA pairs yields a face-specialized MLLM with the highest reported accuracy among MLLMs on FaceXBench.

Reference graph

Works this paper leans on

36 extracted references · 32 canonical work pages · cited by 2 Pith papers

  1. [1]

    Deep Transfer Learning on the Aggre- gated Dataset for Face Presentation Attack Detection

    Faseela Abdullakutty, Eyad Elyan, Pamela Johnston, and Adamu Ali-Gombe. Deep Transfer Learning on the Aggre- gated Dataset for Face Presentation Attack Detection. Cog- nitive Computation, 14(6):2223–2233, Nov. 2022. 3

    work page 2022

  2. [2]

    Critical features for face recognition

    Naphtali Abudarham, Lior Shkiller, and Galit Yovel. Critical features for face recognition. Cognition, 182:73–83, 2019. 3

    work page 2019

  3. [3]

    Gpt-4 technical report, 2024

    Josh Achiam, Steven Adler, Sandhini Agarwal, and Lama Ahmad et al. Gpt-4 technical report, 2024. 1, 2

    work page 2024

  4. [4]

    Chatgpt appli- cations in academic research: A review of benefits, concerns, and recommendations

    Adhari AlZaabi, Amira ALAmri, Halima Albalushi, Ruqaya Aljabri, and AbdulRahman AalAbdulsalam. Chatgpt appli- cations in academic research: A review of benefits, concerns, and recommendations. bioRxiv, 2023. 1

    work page 2023

  5. [5]

    Shefali Arora, M. P. S. Bhatia, and Vipul Mittal. A robust framework for spoofing detection in faces using deep learn- ing. The Visual Computer, 38(7):2461–2472, July 2022. 3

    work page 2022

  6. [6]

    Recent Advances in Face Presentation At- tack Detection, pages 207–228

    Sushil Bhattacharjee, Amir Mohammadi, Andr ´e Anjos, and S´ebastien Marcel. Recent Advances in Face Presentation At- tack Detection, pages 207–228. Springer International Pub- lishing, Cham, 2019. 3

    work page 2019

  7. [7]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Sub- biah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakan- tan, Pranav Shyam, Girish Sastry, Amanda Askell, Sand- hini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz...

    2020

  8. [8]

    On the effectiveness of local binary patterns in face anti- spoofing

    Ivana Chingovska, Andr ´e Anjos, and S ´ebastien Marcel. On the effectiveness of local binary patterns in face anti- spoofing. In 2012 BIOSIG - Proceedings of the International Conference of Biometrics Special Interest Group (BIOSIG) , pages 1–7, 2012. 3

    work page 2012

Show all 36 references
  1. [9]

    Face Recognition Systems Under Spoofing Attacks, pages 165–194

    Ivana Chingovska, Nesli Erdogmus, Andr ´e Anjos, and S´ebastien Marcel. Face Recognition Systems Under Spoofing Attacks, pages 165–194. Springer International Publishing, Cham, 2016. 3

    2016

  2. [10]

    How good is chatgpt at face biometrics? a first look into recognition, soft biometrics, and explainabil- ity

    Ivan Deandres-Tame, Ruben Tolosana, Ruben Vera- Rodriguez, Aythami Morales, Julian Fierrez, and Javier Ortega-Garcia. How good is chatgpt at face biometrics? a first look into recognition, soft biometrics, and explainabil- ity. IEEE Access, 12:34390–34401, 2024. 2

    2024

  3. [11]

    Debayan Deb and Anil K. Jain. Look locally infer globally: A generalizable face anti-spoofing approach. IEEE Transac- tions on Information Forensics and Security, 16:1143–1157,

  4. [12]

    Chatgpt meets iris bio- metrics, 2024

    Parisa Farmanifard and Arun Ross. Chatgpt meets iris bio- metrics, 2024. 2

    2024

  5. [13]

    Bio- metric antispoofing methods: A survey in face recognition

    Javier Galbally, S ´ebastien Marcel, and Julian Fierrez. Bio- metric antispoofing methods: A survey in face recognition. IEEE Access, 2:1530–1552, 01 2014. 3

    2014

  6. [14]

    Deep pixel-wise bi- nary supervision for face presentation attack detection

    Anjith George and S ´ebastien Marcel. Deep pixel-wise bi- nary supervision for face presentation attack detection. In International Conference on Biometrics, 2019. 2, 3, 6

    2019

  7. [15]

    On the effectiveness of vision transformers for zero-shot face anti-spoofing

    Anjith George and S ´ebastien Marcel. On the effectiveness of vision transformers for zero-shot face anti-spoofing. In International Joint Conference on Biometrics (IJCB 2021) ,

    2021

  8. [16]

    Bio- metric face presentation attack detection with multi-channel convolutional neural network

    Anjith George, Zohreh Mostaani, David Geissenbuhler, Olegs Nikisins, Andr ´e Anjos, and S ´ebastien Marcel. Bio- metric face presentation attack detection with multi-channel convolutional neural network. IEEE transactions on infor- mation forensics and security, 15:42–55, 2019. 3

    2019

  9. [17]

    Face anti-spoofing using texture-based techniques and filtering methods

    Md Rezwan Hasan, S M Hasan Mahmud, and Xiang Yu Li. Face anti-spoofing using texture-based techniques and filtering methods. Journal of Physics: Conference Series , 1229(1):012044, may 2019. 3

    2019

  10. [18]

    Chatgpt and biometrics: an assessment of face recognition, gender detec- tion, and age estimation capabilities, 2024

    Ahmad Hassanpour, Yasamin Kowsari, Hatef Otroshi Shahreza, Bian Yang, and Sebastien Marcel. Chatgpt and biometrics: an assessment of face recognition, gender detec- tion, and age estimation capabilities, 2024. 1, 2

    2024

  11. [19]

    Hernandez-Ortega, J

    J. Hernandez-Ortega, J. Fierrez, A. Morales, and J. Gal- bally. Introduction to Presentation Attack Detection in Face Biometrics and Recent Advances , chapter in Handbook of Biometric Anti-Spoofing: Presentation Attack Detection and Vulnerability Assessment, page 203–230. Spri...

  12. [20]

    Can chatgpt detect deepfakes? a study of using mul- timodal large language models for media forensics

    Shan Jia, Reilin Lyu, Kangran Zhao, Yize Chen, Zhiyuan Yan, Yan Ju, Chuanbo Hu, Xin Li, Baoyuan Wu, and Siwei Lyu. Can chatgpt detect deepfakes? a study of using mul- timodal large language models for media forensics. In Pro- ceedings of the IEEE/CVF Conference on Computer Vis...

    2024

  13. [21]

    Face spoofing and counter-spoofing: A survey of state-of-the-art algorithms

    Dakshina Ranjan Kisku and Rinku Datta Rakshit. Face spoofing and counter-spoofing: A survey of state-of-the-art algorithms. Transactions on Engineering and Computing Sciences, 5(2):31, May 2017. 3

    2017

  14. [22]

    2d face liveness detection: An overview

    Olga K ¨ahm and Naser Damer. 2d face liveness detection: An overview. In2012 BIOSIG - Proceedings of the International Conference of Biometrics Special Interest Group (BIOSIG) , pages 1–12, 2012. 3

    2012

  15. [23]

    Self-supervised 2d face presentation attack detection via temporal sequence sampling

    Usman Muhammad, Zitong Yu, and Jukka Komulainen. Self-supervised 2d face presentation attack detection via temporal sequence sampling. Pattern Recognition Letters , 156:15–22, 2022. 3

    2022

  16. [24]

    F. Peng, L. Qin, and M. Long. Face presentation attack de- tection using guided scale texture. Multimed Tools Appl , 77:8883–8909, 2018. 3

    2018

  17. [25]

    Learning transferable visual models from natural language supervision, 2021

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision, 2021. 3

    2021

  18. [26]

    Presenta- tion attack detection methods for face recognition systems: A comprehensive survey

    Raghavendra Ramachandra and Christoph Busch. Presenta- tion attack detection methods for face recognition systems: A comprehensive survey. ACM Comput. Surv., 50(1), Mar

  19. [27]

    A novel and responsible dataset for face presentation attack de- tection on mobile devices

    Nathan Ramoly, Alain Komaty, Vedrana Krivokuca, Lara Younes, Ahmad-Montaser Awal, and S ´ebastien Marcel. A novel and responsible dataset for face presentation attack de- tection on mobile devices. In The IEEE International Joint Conference on Biometrics, page 8, 2024. 3

    2024

  20. [28]

    A survey on face pre- sentation attack detection mechanisms: hitherto and future perspectives

    Deepika Sharma and Arvind Selwal. A survey on face pre- sentation attack detection mechanisms: hitherto and future perspectives. Multimedia Systems, 29(3):1527–1577, June

  21. [29]

    Shield : An evaluation benchmark for face spoofing and forgery detection with multimodal large lan- guage models

    Yichen Shi, Yuhao Gao, Yingxin Lai, Hongyang Wang, Jun Feng, Lei He, Jun Wan, Changsheng Chen, Zitong Yu, and Xiaochun Cao. Shield : An evaluation benchmark for face spoofing and forgery detection with multimodal large lan- guage models. ArXiv, abs/2402.04178, 2024. 2, 3

    2024 arXiv

  22. [30]

    How far did we get in face spoofing detection? Engi- neering Applications of Artificial Intelligence , 72:368–381,

    Luiz Souza, Luciano Oliveira, Mauricio Pamplona, and Joao Papa. How far did we get in face spoofing detection? Engi- neering Applications of Artificial Intelligence , 72:368–381,

  23. [31]

    Flip: Cross-domain face anti-spoofing with lan- guage guidance, 2023

    Koushik Srivatsan, Muzammal Naseer, and Karthik Nan- dakumar. Flip: Cross-domain face anti-spoofing with lan- guage guidance, 2023. 3

    2023

  24. [32]

    Chatgpt in medicine: an overview of its applications, advantages, limitations, future prospects, and ethical considerations

    JDave T, Athaluri SA, and Singh S. Chatgpt in medicine: an overview of its applications, advantages, limitations, future prospects, and ethical considerations. In Front Artif Intell, May 2023. 1

    2023

  25. [33]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need, 2023. 2

    2023

  26. [34]

    Image analysis through the lens of chatgpt-

    Johnson Victor, Osamah Alyasiri, Dua’A Akhtom, and Olabisi Johnson. Image analysis through the lens of chatgpt-

  27. [35]

    Journal of Applied Artificial Intelligence , 4:32–46, 12

  28. [36]

    A learnable gra- dient operator for face presentation attack detection

    Caixun Wang, Bingyao Yu, and Jie Zhou. A learnable gra- dient operator for face presentation attack detection. Pattern Recognition, 135:109146, 2023. 3

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.