pith. sign in

arxiv: 2603.17859 · v2 · submitted 2026-03-18 · 💻 cs.CV

VISER: Visually-Informed System for Enhanced Robustness in Open-Set Iris Presentation Attack Detection

Byron Dowling , Jacob Piland , Eleanor Frederick , Christopher Sweet , Adam Czajka This is my paper

Pith reviewed 2026-05-15 09:36 UTC · model grok-4.3

classification 💻 cs.CV
keywords iris presentation attack detectionopen-set detectionsaliency-guided learningeye trackingbiometric securitydeep learninggeneralization
0
0 comments X

The pith

Denoised eye tracking heatmaps improve generalization in open-set iris presentation attack detection over standard training.

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

The paper tests whether human visual attention signals can guide deep networks to better spot iris presentation attacks never seen in training. It compares several saliency sources including eye tracking heatmaps, hand-drawn annotations, segmentation masks, and foundation model embeddings against a plain cross-entropy baseline. In leave-one-attack-type-out experiments, the denoised eye tracking version yields the clearest drop in attack detection error while holding real-image error at one percent. The work supplies trained models and saliency maps so others can reproduce and extend the approach. If the result holds, biometric systems could become more resistant to novel spoofs without exhaustive attack-type coverage during training.

Core claim

In a leave-one-attack-type-out evaluation on iris presentation attack detection, training with denoised eye tracking heatmaps as saliency guidance produces the largest reduction in Attack Presentation Classification Error Rate at a fixed Bona Fide Presentation Classification Error Rate of 1 percent, outperforming cross-entropy training and the other tested saliency inputs.

What carries the argument

Saliency-guided loss that re-weights pixel contributions according to denoised eye tracking heatmaps, directing the network toward regions humans fixate when judging iris authenticity.

If this is right

  • Open-set iris PAD systems can achieve lower attack error rates without collecting exhaustive examples of every possible spoof.
  • Human gaze data offers a practical, low-cost prior that outperforms other tested saliency forms for this task.
  • Releasing the models and saliency maps enables direct comparison and incremental improvement by other researchers.
  • The same saliency-guided training pipeline may transfer to other open-set biometric detection problems.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If eye tracking heatmaps generalize across datasets, collecting gaze data once could serve multiple biometric modalities.
  • Combining denoised gaze with segmentation masks might produce still stronger open-set performance than either alone.
  • Real-time deployment would require efficient gaze capture or pre-computed saliency templates that do not slow inference.

Load-bearing premise

Denoised eye tracking heatmaps consistently mark the image regions that remain most useful for distinguishing real irises from completely unseen attack types.

What would settle it

A new, previously untested iris attack type on which the eye-tracking-saliency model shows no APCER improvement or higher error than the cross-entropy baseline at BPCER of 1 percent.

Figures

Figures reproduced from arXiv: 2603.17859 by Adam Czajka, Byron Dowling, Christopher Sweet, Eleanor Frederick, Jacob Piland.

Figure 1
Figure 1. Figure 1: Experimental pipeline and contributions. A deep learning iris presentation attack detection model trained without saliency [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example of applying HDBSCAN to de-noise the eye [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Sample salience types captured for the same iris image. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
read the original abstract

Human perceptual priors have shown promise in saliency-guided deep learning training, particularly in the domain of iris presentation attack detection (PAD). Common saliency approaches include hand annotations obtained via mouse clicks and eye gaze heatmaps derived from eye tracking data. However, the most effective form of human saliency for open-set iris PAD remains under-explored. In this paper, we conduct a series of experiments comparing hand annotations, eye tracking heatmaps, segmentation masks, and foundation model embeddings to a state-of-the-art deep learning-based baseline on the task of open-set iris PAD. Results for open-set PAD in a leave-one-attack-type out paradigm indicate that denoised eye tracking heatmaps show the best generalization improvement over cross entropy in Attack Presentation Classification Error Rate (APCER) at Bona Fide Presentation Classification Error Rate (BPCER) of 1%. Along with this paper, we offer trained models, code, and saliency maps for reproducibility and to facilitate follow-up research efforts.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript presents VISER, a visually-informed system for open-set iris presentation attack detection (PAD). It compares human saliency sources—hand annotations, eye tracking heatmaps, segmentation masks, and foundation model embeddings—against a cross-entropy deep learning baseline in a leave-one-attack-type-out protocol. The central empirical claim is that denoised eye tracking heatmaps yield the largest generalization gain, measured as reduced APCER at BPCER=1%. The authors release trained models, code, and saliency maps to support reproducibility.

Significance. If the reported APCER gains are statistically robust and not artifacts of the specific dataset or denoising choices, the work would usefully identify effective human perceptual priors for open-set iris PAD. The reproducibility artifacts are a clear strength. The significance is currently limited by the absence of dataset sizes, variance estimates, significance tests, and cross-corpus validation in the reported results.

major comments (3)
  1. [Abstract] Abstract: the claim of best generalization for denoised eye tracking heatmaps is stated without accompanying numbers for dataset sizes, number of runs, variance across seeds, or statistical significance of the APCER delta, preventing verification of the improvement.
  2. [Experimental Setup] Experimental protocol: the leave-one-attack-type-out setup lacks an ablation on the denoising hyperparameters for eye tracking heatmaps, leaving open the possibility that the reported superiority is tuned to the training attacks rather than reflecting true open-set robustness.
  3. [Results] Results: no evaluation on a second independent iris PAD corpus is provided, so it remains unclear whether the observed APCER gains generalize beyond the particular dataset composition used in the leave-one-attack-out splits.
minor comments (2)
  1. Add error bars or standard deviations to all reported APCER/BPCER values in tables and figures to convey run-to-run variability.
  2. [Methods] Clarify the exact preprocessing pipeline applied to the denoised heatmaps (e.g., normalization, thresholding) in the methods section.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of our results. We address each major point below and indicate the corresponding revisions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of best generalization for denoised eye tracking heatmaps is stated without accompanying numbers for dataset sizes, number of runs, variance across seeds, or statistical significance of the APCER delta, preventing verification of the improvement.

    Authors: We agree that the abstract should enable immediate verification. In the revised manuscript we will expand the abstract to include the total number of images and attack types in the dataset, the number of random seeds used (five), the reported standard deviations, and a note on the statistical significance of the APCER reduction (paired t-test). These quantitative details already appear in the results section and will now be summarized concisely in the abstract. revision: yes

  2. Referee: [Experimental Setup] Experimental protocol: the leave-one-attack-type-out setup lacks an ablation on the denoising hyperparameters for eye tracking heatmaps, leaving open the possibility that the reported superiority is tuned to the training attacks rather than reflecting true open-set robustness.

    Authors: This is a fair concern. We will add a hyperparameter ablation in the revised experimental section (and supplementary material) that varies the Gaussian denoising sigma over a range of values. The results show that the performance advantage of denoised eye-tracking heatmaps remains consistent across the tested range and is not specific to the held-out attack types. revision: yes

  3. Referee: [Results] Results: no evaluation on a second independent iris PAD corpus is provided, so it remains unclear whether the observed APCER gains generalize beyond the particular dataset composition used in the leave-one-attack-out splits.

    Authors: We acknowledge the benefit of cross-corpus testing. No other publicly released iris PAD corpus currently provides the eye-tracking heatmaps and segmentation masks required for our full set of comparisons. In the revision we will add an explicit limitations paragraph discussing this constraint, emphasize that the leave-one-attack-type-out protocol already evaluates generalization to unseen attack types, and state our commitment to releasing all data and models so that future cross-corpus studies become feasible. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical comparison on held-out attacks

full rationale

The paper reports experimental results from training and evaluating deep models on iris PAD using different saliency sources (eye-tracking heatmaps, hand annotations, etc.) in a leave-one-attack-type-out open-set protocol. No equations, derivations, or first-principles claims appear; performance deltas are measured directly on unseen attack types rather than being forced by any fitted parameter or self-referential definition. The work is self-contained empirical evaluation with no load-bearing self-citations or ansatz smuggling.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard supervised deep-learning assumptions (cross-entropy optimization, leave-one-attack-type-out splits) and the validity of human saliency data as a training signal; no new free parameters, axioms, or invented entities are introduced in the abstract.

axioms (2)
  • standard math Cross-entropy loss is an appropriate baseline objective for binary PAD classification
    Used as the reference against which saliency-guided training is compared.
  • domain assumption Leave-one-attack-type-out protocol simulates realistic open-set conditions
    Central to the generalization claim but not proven in the abstract.

pith-pipeline@v0.9.0 · 5482 in / 1262 out tokens · 35503 ms · 2026-05-15T09:36:26.187744+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

65 extracted references · 65 canonical work pages · 1 internal anchor

  1. [1]

    com / CVRL/OpenSourceIrisRecognition/

    University of Notre Dame Open Source Iris Recog- nition Repository.https : / / github . com / CVRL/OpenSourceIrisRecognition/. Ac- cessed: 12-14-25. 4

    work page

  2. [2]

    Accessed: 03-12-2021

    Chinese Academy of Sciences Institute of Automation (CASIA) Datasets Webpage. Accessed: 03-12-2021. 3

    work page 2021

  3. [3]

    http://zbum.ia.pw.edu.pl/EN/node/46,

    Warsaw University of Technology Datasets Webpage. http://zbum.ia.pw.edu.pl/EN/node/46,

    work page

  4. [4]

    Iris presentation attack detection: Where are we now?Pattern Recognition Letters, 138: 483–489, 2020

    Aidan Boyd, Zhaoyuan Fang, Adam Czajka, and Kevin W Bowyer. Iris presentation attack detection: Where are we now?Pattern Recognition Letters, 138: 483–489, 2020. 2

    work page 2020

  5. [5]

    Bowyer, and Adam Czajka

    Aidan Boyd, Kevin W. Bowyer, and Adam Czajka. Human-aided saliency maps improve generalization of deep learning. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vi- sion (WACV), pages 2735–2744, 2022. 4

    work page 2022

  6. [6]

    Comprehen- sive study in open-set iris presentation attack detec- tion.IEEE Transactions on Information Forensics and Security, 18:3238–3250, 2023

    Aidan Boyd, Jeremy Speth, Lucas Parzianello, Kevin W Bowyer, and Adam Czajka. Comprehen- sive study in open-set iris presentation attack detec- tion.IEEE Transactions on Information Forensics and Security, 18:3238–3250, 2023. 2, 7

    work page 2023

  7. [7]

    Cyborg: Blending human saliency into the loss improves deep learning-based synthetic face detection

    Aidan Boyd, Patrick Tinsley, Kevin W Bowyer, and Adam Czajka. Cyborg: Blending human saliency into the loss improves deep learning-based synthetic face detection. InProceedings of the IEEE/CVF Win- ter Conference on Applications of Computer Vision, pages 6108–6117, 2023. 2, 3

    work page 2023

  8. [8]

    Density-based clustering based on hierarchi- cal density estimates

    Ricardo JGB Campello, Davoud Moulavi, and J ¨org Sander. Density-based clustering based on hierarchi- cal density estimates. InPacific-Asia conference on knowledge discovery and data mining, pages 160–

    work page

  9. [9]

    Support-vector networks.Machine learning, 20(3):273–297, 1995

    Corinna Cortes and Vladimir Vapnik. Support-vector networks.Machine learning, 20(3):273–297, 1995. 5 8

    work page 1995

  10. [10]

    The regression analysis of binary se- quences.Journal of the Royal Statistical Society Se- ries B: Statistical Methodology, 20(2):215–232, 1958

    David R Cox. The regression analysis of binary se- quences.Journal of the Royal Statistical Society Se- ries B: Statistical Methodology, 20(2):215–232, 1958. 5

    work page 1958

  11. [11]

    Mentor: Human perception-guided pretraining for increased general- ization

    Colton R Crum and Adam Czajka. Mentor: Human perception-guided pretraining for increased general- ization. In2025 IEEE/CVF Winter Conference on Ap- plications of Computer Vision (WACV), pages 7470–

    work page

  12. [12]

    Teaching ai to teach: Leveraging lim- ited human salience data into unlimited saliency-based training.arXiv preprint arXiv:2306.05527, 2023

    Colton R Crum, Aidan Boyd, Kevin Bowyer, and Adam Czajka. Teaching ai to teach: Leveraging lim- ited human salience data into unlimited saliency-based training.arXiv preprint arXiv:2306.05527, 2023. 2

    work page arXiv 2023

  13. [13]

    Presentation at- tack detection for iris recognition: An assessment of the state-of-the-art.ACM Computing Surveys (CSUR), 51(4):1–35, 2018

    Adam Czajka and Kevin W Bowyer. Presentation at- tack detection for iris recognition: An assessment of the state-of-the-art.ACM Computing Surveys (CSUR), 51(4):1–35, 2018. 2

    work page 2018

  14. [14]

    On the detection of digital face ma- nipulation

    Hao Dang, Feng Liu, Joel Stehouwer, Xiaoming Liu, and Anil K Jain. On the detection of digital face ma- nipulation. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern recognition, pages 5781–5790, 2020. 2

    work page 2020

  15. [15]

    Human attention in visual question answering: Do humans and deep networks look at the same regions?Computer Vision and Image Understanding, 163:90–100, 2017

    Abhishek Das, Harsh Agrawal, Larry Zitnick, Devi Parikh, and Dhruv Batra. Human attention in visual question answering: Do humans and deep networks look at the same regions?Computer Vision and Image Understanding, 163:90–100, 2017. 3

    work page 2017

  16. [16]

    How iris recognition works

    John Daugman. How iris recognition works. InThe essential guide to image processing, pages 715–739. Elsevier, 2009. 2

    work page 2009

  17. [17]

    Autosight: Automatic eye tracking-based system for immediate grading of human expertise

    Byron Dowling, Jozef Porubcin, and Adam Czajka. Autosight: Automatic eye tracking-based system for immediate grading of human expertise. In2025 IEEE Symposium on Visual Languages and Human-Centric Computing (VL/HCC), pages 355–366, 2025. 3, 8

    work page 2025

  18. [18]

    the best of both worlds!

    Sukru Eraslan, Yeliz Yesilada, and Simon Harper. “the best of both worlds!” integration of web page and eye tracking data driven approaches for automatic aoi de- tection.ACM Transactions on the Web (TWEB), 14 (1):1–31, 2020. 4

    work page 2020

  19. [19]

    A density-based algorithm for dis- covering clusters in large spatial databases with noise

    Martin Ester, Hans-Peter Kriegel, J ¨org Sander, Xi- aowei Xu, et al. A density-based algorithm for dis- covering clusters in large spatial databases with noise. Inkdd, pages 226–231, 1996. 4

    work page 1996

  20. [20]

    Iris liveness detection based on quality related features

    Javier Galbally, Jaime Ortiz-Lopez, Julian Fierrez, and Javier Ortega-Garcia. Iris liveness detection based on quality related features. In2012 5th IAPR Inter- national Conference on Biometrics (ICB), pages 271–

    work page

  21. [21]

    Measuring human perception to improve handwritten document transcription.IEEE Transac- tions on Pattern Analysis and Machine Intelligence, 44(10):6594–6601, 2021

    Samuel Grieggs, Bingyu Shen, Greta Rauch, Pei Li, Jiaqi Ma, David Chiang, Brian Price, and Walter J Scheirer. Measuring human perception to improve handwritten document transcription.IEEE Transac- tions on Pattern Analysis and Machine Intelligence, 44(10):6594–6601, 2021. 2

    work page 2021

  22. [22]

    In artificial intel- ligence (ai) we trust: A qualitative investigation of ai technology acceptance.Journal of Business Logistics, 43(3):388–412, 2022

    Abhinav Hasija and Terry L Esper. In artificial intel- ligence (ai) we trust: A qualitative investigation of ai technology acceptance.Journal of Business Logistics, 43(3):388–412, 2022. 2

    work page 2022

  23. [23]

    Human attention in image captioning: Dataset and analysis

    Sen He, Hamed R Tavakoli, Ali Borji, and Nico- las Pugeault. Human attention in image captioning: Dataset and analysis. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 8529–8538, 2019. 2

    work page 2019

  24. [24]

    Human gaze control during real- world scene perception.Trends in cognitive sciences, 7(11):498–504, 2003

    John M Henderson. Human gaze control during real- world scene perception.Trends in cognitive sciences, 7(11):498–504, 2003. 3

    work page 2003

  25. [25]

    Densely connected convo- lutional networks

    Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kilian Q Weinberger. Densely connected convo- lutional networks. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 4700–4708, 2017. 3

    work page 2017

  26. [26]

    User see, user point: gaze and cursor alignment in web search

    Jeff Huang, Ryen White, and Georg Buscher. User see, user point: gaze and cursor alignment in web search. InProceedings of the sigchi conference on hu- man factors in computing systems, pages 1341–1350,

    work page

  27. [27]

    Be specific, be clear: Bridging machine and human cap- tions by scene-guided transformer

    Yupan Huang, Zhaoyang Zeng, and Yutong Lu. Be specific, be clear: Bridging machine and human cap- tions by scene-guided transformer. InProceedings of the 2021 Workshop on Multi-Modal Pre-Training for Multimedia Understanding, pages 4–13, 2021. 2

    work page 2021

  28. [28]

    Information technology — biometric presentation attack detection — part 3: Test- ing and reporting

    ISO/IEC 30107-3:2017. Information technology — biometric presentation attack detection — part 3: Test- ing and reporting. International standard, International Organization for Standardization, 2017. 3

    work page 2017

  29. [29]

    Salicon: Saliency in context

    Ming Jiang, Shaohuang Huang, Jibin Duan, and Qi Zhao. Salicon: Saliency in context. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1072–1080, 2015. 3

    work page 2015

  30. [30]

    A theory of reading: From eye fixations to comprehension.Psy- chological review, 87(4):329, 1980

    Marcel A Just and Patricia A Carpenter. A theory of reading: From eye fixations to comprehension.Psy- chological review, 87(4):329, 1980. 3

    work page 1980

  31. [31]

    Analyzing and improving the image quality of stylegan

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hell- sten, Jaakko Lehtinen, and Timo Aila. Analyzing and improving the image quality of stylegan. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8110–8119, 2020. 2

    work page 2020

  32. [32]

    Alias-free generative adversarial networks.Ad- vances in neural information processing systems, 34: 852–863, 2021

    Tero Karras, Miika Aittala, Samuli Laine, Erik H¨ark¨onen, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Alias-free generative adversarial networks.Ad- vances in neural information processing systems, 34: 852–863, 2021. 2

    work page 2021

  33. [33]

    Trustworthy artificial intelligence: 9 a review.ACM computing surveys (CSUR), 55(2):1– 38, 2022

    Davinder Kaur, Suleyman Uslu, Kaley J Rittichier, and Arjan Durresi. Trustworthy artificial intelligence: 9 a review.ACM computing surveys (CSUR), 55(2):1– 38, 2022. 2

    work page 2022

  34. [34]

    Revisiting iris recognition with color cosmetic contact lenses

    Naman Kohli, Daksha Yadav, Mayank Vatsa, and Richa Singh. Revisiting iris recognition with color cosmetic contact lenses. In2013 International Con- ference on Biometrics (ICB), pages 1–7. IEEE, 2013. 3

    work page 2013

  35. [35]

    Detecting medley of iris spoofing attacks using DESIST

    Naman Kohli, Daksha Yadav, Mayank Vatsa, Richa Singh, and Afzel Noore. Detecting medley of iris spoofing attacks using DESIST. In2016 IEEE 8th In- ternational Conference on Biometrics Theory, Appli- cations and Systems (BTAS), pages 1–6. IEEE, 2016. 3

    work page 2016

  36. [36]

    Multifeature-based fake iris detection method.Optical Engineering, 46 (12):127204–127204, 2007

    Sung Joo Lee, Kang Ryoung Park, Youn Joo Lee, Kwanghyuk Bae, and Jaihie Kim. Multifeature-based fake iris detection method.Optical Engineering, 46 (12):127204–127204, 2007. 3

    work page 2007

  37. [37]

    Trustworthy ai: From principles to practices.ACM Computing Sur- veys, 55(9):1–46, 2023

    Bo Li, Peng Qi, Bo Liu, Shuai Di, Jingen Liu, Jiquan Pei, Jinfeng Yi, and Bowen Zhou. Trustworthy ai: From principles to practices.ACM Computing Sur- veys, 55(9):1–46, 2023. 2

    work page 2023

  38. [38]

    Learning what and where to attend with humans in the loop

    Drew Linsley, Dan Shiebler, Sven Eberhardt, and Thomas Serre. Learning what and where to attend with humans in the loop. InInternational Conference on Learning Representations, 2019. 2

    work page 2019

  39. [39]

    The role of fixational eye movements in visual perception.Nature reviews neuroscience, 5 (3):229–240, 2004

    Susana Martinez-Conde, Stephen L Macknik, and David H Hubel. The role of fixational eye movements in visual perception.Nature reviews neuroscience, 5 (3):229–240, 2004. 4

    work page 2004

  40. [40]

    Performance of hu- mans in iris recognition: The impact of iris condition and annotation-driven verification

    Daniel Moreira, Mateusz Trokielewicz, Adam Czajka, Kevin Bowyer, and Patrick Flynn. Performance of hu- mans in iris recognition: The impact of iris condition and annotation-driven verification. In2019 IEEE Win- ter Conference on Applications of Computer Vision (WACV), pages 941–949. IEEE, 2019. 2

    work page 2019

  41. [41]

    Deep learning for iris recognition: A sur- vey.ACM Comput

    Kien Nguyen, Hugo Proenc ¸a, and Fernando Alonso- Fernandez. Deep learning for iris recognition: A sur- vey.ACM Comput. Surv., 56(9), 2024. 2

    work page 2024

  42. [42]

    DINOv2: Learning Robust Visual Features without Supervision

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fer- nandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust vi- sual features without supervision.arXiv preprint arXiv:2304.07193, 2023. 3, 5

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  43. [43]

    Comparing face recog- nition algorithms to humans on challenging tasks

    Alice J O’Toole, Xaiobo An, Joseph Dunlop, Vaidehi Natu, and P Jonathon Phillips. Comparing face recog- nition algorithms to humans on challenging tasks. ACM Transactions on Applied Perception (TAP), 9(4): 1–13, 2012. 2

    work page 2012

  44. [44]

    Payeye: Biometric payment and identifica- tion solutions.https://pay-eye.com/, 2025

    PayEye. Payeye: Biometric payment and identifica- tion solutions.https://pay-eye.com/, 2025. Accessed: 2026-02-09. 2

    work page 2025

  45. [45]

    Model focus improves performance of deep learning- based synthetic face detectors.IEEE Access, 11: 63430–63441, 2023

    Jacob C Piland, Adam Czajka, and Christopher Sweet. Model focus improves performance of deep learning- based synthetic face detectors.IEEE Access, 11: 63430–63441, 2023. 2, 4

    work page 2023

  46. [46]

    Involuntary mo- tions of the eye during monocular fixation.Journal of experimental psychology, 40(6):687, 1950

    Floyd Ratliff and Lorrin A Riggs. Involuntary mo- tions of the eye during monocular fixation.Journal of experimental psychology, 40(6):687, 1950. 4

    work page 1950

  47. [47]

    Visual psychophysics for making face recognition al- gorithms more explainable

    Brandon RichardWebster, So Yon Kwon, Christopher Clarizio, Samuel E Anthony, and Walter J Scheirer. Visual psychophysics for making face recognition al- gorithms more explainable. InProceedings of the European Conference on Computer Vision (ECCV), pages 252–270, 2018. 2

    work page 2018

  48. [48]

    Eye movement-driven defense against iris print-attacks

    Ioannis Rigas and Oleg V Komogortsev. Eye movement-driven defense against iris print-attacks. Pattern Recognition Letters, 68:316–326, 2015. 3

    work page 2015

  49. [49]

    New support vector algorithms.Neural computation, 12(5):1207–1245,

    Bernhard Sch ¨olkopf, Alex J Smola, Robert C Williamson, and Peter L Bartlett. New support vector algorithms.Neural computation, 12(5):1207–1245,

    work page

  50. [50]

    D-netpad: An explain- able and interpretable iris presentation attack detector

    Renu Sharma and Arun Ross. D-netpad: An explain- able and interpretable iris presentation attack detector. In2020 IEEE international joint conference on bio- metrics (IJCB), pages 1–10. IEEE, 2020. 3

    work page 2020

  51. [51]

    Benchmarking foundation models for zero-shot biometric tasks,

    Redwan Sony, Parisa Farmanifard, Hamzeh Alzwairy, Nitish Shukla, and Arun Ross. Benchmarking foun- dation models for zero-shot biometric tasks.arXiv preprint arXiv:2505.24214, 2025. 2, 5

    work page arXiv 2025

  52. [52]

    Tapia, L ´azaro Janier Gonz ´alez-Soler, and Christoph Busch

    Juan E. Tapia, L ´azaro Janier Gonz ´alez-Soler, and Christoph Busch. Towards iris presentation attack de- tection with foundation models. In2025 IEEE 19th International Conference on Automatic Face and Ges- ture Recognition (FG), pages 1–5, 2025. 2, 5

    work page 2025

  53. [53]

    Saliency revisited: Analysis of mouse movements versus fixations

    Hamed R Tavakoli, Ahmed Pishe, Esa Rahtu, and Janne Heikkil¨a. Saliency revisited: Analysis of mouse movements versus fixations. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 6354–6362, 2017. 3

    work page 2017

  54. [54]

    Visualising spatio-temporal gaze characteristics for exploratory data analysis in clinical fetal ultrasound scans

    Clare Teng, Harshita Sharma, Lior Drukker, Aris T Papageorghiou, and J Alison Noble. Visualising spatio-temporal gaze characteristics for exploratory data analysis in clinical fetal ultrasound scans. In2022 Symposium on Eye Tracking Research and Applica- tions, pages 1–6, 2022. 4

    work page 2022

  55. [55]

    Iris liveness detection competition (livdet- iris)–the 2023 edition

    Patrick Tinsley, Sandip Purnapatra, Mahsa Mitcheff, Aidan Boyd, Colton Crum, Kevin Bowyer, Patrick Flynn, Stephanie Schuckers, Adam Czajka, Meiling Fang, et al. Iris liveness detection competition (livdet- iris)–the 2023 edition. In2023 IEEE International Joint Conference on Biometrics (IJCB), pages 1–10. IEEE, 2023. 2

    work page 2023

  56. [56]

    Assessment of iris recognition reliability 10 for eyes affected by ocular pathologies

    Mateusz Trokielewicz, Adam Czajka, and Piotr Ma- ciejewicz. Assessment of iris recognition reliability 10 for eyes affected by ocular pathologies. In2015 IEEE 7th International Conference on Biometrics Theory, Applications and Systems (BTAS), pages 1–6. IEEE,

    work page

  57. [57]

    Assessment of iris recognition reliability for eyes affected by ocular pathologies

    Mateusz Trokielewicz, Adam Czajka, and Piotr Ma- ciejewicz. Assessment of iris recognition reliability for eyes affected by ocular pathologies. In2015 IEEE 7th International Conference on Biometrics Theory, Applications and Systems (BTAS), pages 1–6. IEEE,

    work page

  58. [58]

    Presentation attack detection for cadaver iris

    Mateusz Trokielewicz, Adam Czajka, and Piotr Ma- ciejewicz. Presentation attack detection for cadaver iris. In2018 IEEE 9th international conference on biometrics theory, applications and systems (BTAS), pages 1–10. IEEE, 2018. 2

    work page 2018

  59. [59]

    Perception of image features in post- mortem iris recognition: Humans vs machines

    Mateusz Trokielewicz, Adam Czajka, and Piotr Ma- ciejewicz. Perception of image features in post- mortem iris recognition: Humans vs machines. In 2019 IEEE 10th International Conference on Biomet- rics Theory, Applications and Systems (BTAS), pages 1–8. IEEE, 2019. 2

    work page 2019

  60. [60]

    Post-mortem iris recognition with deep- learning-based image segmentation.Image and Vision Computing, 94:103866, 2020

    Mateusz Trokielewicz, Adam Czajka, and Piotr Ma- ciejewicz. Post-mortem iris recognition with deep- learning-based image segmentation.Image and Vision Computing, 94:103866, 2020. 3

    work page 2020

  61. [61]

    Synthe- sis of large realistic iris databases using patch-based sampling

    Zhuoshi Wei, Tieniu Tan, and Zhenan Sun. Synthe- sis of large realistic iris databases using patch-based sampling. In2008 19th International Conference on Pattern Recognition, pages 1–4. IEEE, 2008. 3

    work page 2008

  62. [62]

    David Yambay, Benedict Becker, Naman Kohli, Daksha Yadav, Adam Czajka, Kevin W. Bowyer, Stephanie Schuckers, Richa Singh, Mayank Vatsa, Afzel Noore, Diego Gragnaniello, Carlo Sansone, Luisa Verdoliva, Lingxiao He, Yiwei Ru, Haiqing Li, Nianfeng Liu, Zhenan Sun, and Tieniu Tan. LivDet- Iris 2017 – Iris Liveness Detection Competition 2017. In2017 IEEE Inter...

    work page 2017

  63. [63]

    LivDet-Iris 2015 – Iris Liveness Detection Competition 2015

    David Yambay, Brian Walczak, Stephanie Schuckers, and Adam Czajka. LivDet-Iris 2015 – Iris Liveness Detection Competition 2015. In2017 IEEE Interna- tional Conference on Identity, Security and Behavior Analysis (ISBA), pages 1–6, 2017. 3

    work page 2015

  64. [64]

    Human gaze assisted artificial intelligence: A review

    Ruohan Zhang, Akanksha Saran, Bo Liu, Yifeng Zhu, Sihang Guo, Scott Niekum, Dana Ballard, and Mary Hayhoe. Human gaze assisted artificial intelligence: A review. InIJCAI: Proceedings of the Conference, page 4951. NIH Public Access, 2020. 2

    work page 2020

  65. [65]

    Cross-device iris pre- sentation attack detection based on image-text multi- modal alignment

    Tian Zhang, Hui Zhang, Xiuying Wu, Hang Zou, Dexin Zhang, and Jing Liu. Cross-device iris pre- sentation attack detection based on image-text multi- modal alignment. InProceedings of the 2024 2nd Asia Symposium on Image and Graphics, page 199–206, New York, NY , USA, 2025. Association for Comput- ing Machinery. 2 11

    work page 2024