{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,18]],"date-time":"2026-11-18T17:52:51Z","timestamp":1795024371679,"version":"4.10.2"},"reference-count":85,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","license":[{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/legalcode"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Access"],"published-print":{"date-parts":[[2025]]},"DOI":"10.1109\/access.2025.3536784","type":"journal-article","created":{"date-parts":[[2025,1,29]],"date-time":"2025-01-29T19:02:57Z","timestamp":1738177377000},"page":"26050-26062","source":"Crossref","is-referenced-by-count":5,"title":["On Privacy, Accuracy, and Fairness Trade-Offs in Facial Recognition"],"prefix":"10.1109","volume":"13","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-0287-513X","authenticated-orcid":false,"given":"Amir","family":"Zarei","sequence":"first","affiliation":[{"name":"Department of Information Security and Communication Technology, Norwegian University of Science and Technology, Gj&#x00F8;vik, Norway"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3936-2223","authenticated-orcid":false,"given":"Ahmad","family":"Hassanpour","sequence":"additional","affiliation":[{"name":"Department of Information Security and Communication Technology, Norwegian University of Science and Technology, Gj&#x00F8;vik, Norway"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9489-5161","authenticated-orcid":false,"given":"Kiran","family":"Raja","sequence":"additional","affiliation":[{"name":"Department of Computer Science, Norwegian University of Science and Technology, Gj&#x00F8;vik, Norway"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.02352"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1145\/3357713.3384290"},{"key":"ref3","article-title":"Generalization techniques empirically outperform differential privacy against membership inference","author":"Liu","year":"2021","journal-title":"arXiv:2110.05524"},{"key":"ref4","article-title":"Efficient sharpness-aware minimization for improved training of neural networks","author":"Du","year":"2021","journal-title":"arXiv:2110.03141"},{"key":"ref5","first-page":"27204","article-title":"Differentially private sharpnessaware training","volume-title":"Proc. PMLR","author":"Park"},{"key":"ref6","first-page":"2615","article-title":"Systematic evaluation of privacy risks of machine learning models","volume-title":"Proc. 30th USENIX Secur. Symp.","author":"Song"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00482"},{"key":"ref9","first-page":"1895","article-title":"Evaluating differentially private machine learning in practice","volume-title":"Proc. 28th USENIX Secur. Symp.","author":"Jayaraman"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/EuroSP51992.2021.00028"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/cvprw50498.2020.00359"},{"key":"ref12","article-title":"Sharpness-aware minimization for efficiently improving generalization","author":"Foret","year":"2020","journal-title":"arXiv:2010.01412"},{"key":"ref13","article-title":"Investigating trade-offs in utility, fairness and differential privacy in neural networks","author":"Pannekoek","year":"2021","journal-title":"arXiv:2102.05975"},{"key":"ref14","article-title":"Differentially private training of residual networks with scale normalisation","author":"Klause","year":"2022","journal-title":"arXiv:2203.00324"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1561\/9781601988195"},{"key":"ref16","first-page":"38305","article-title":"Scalable and efficient training of large convolutional neural networks with differential privacy","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Bu"},{"key":"ref17","article-title":"Unlocking highaccuracy differentially private image classification through scale","author":"De","year":"2022","journal-title":"arXiv:2204.13650"},{"key":"ref18","article-title":"On the importance of difficulty calibration in membership inference attacks","author":"Watson","year":"2021","journal-title":"arXiv:2111.08440"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00819"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/MLSP52302.2021.9596307"},{"key":"ref21","article-title":"Large-scale differentially private BERT","author":"Anil","year":"2021","journal-title":"arXiv:2108.01624"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.findings-emnlp.102"},{"key":"ref23","article-title":"Large language models can be strong differentially private learners","author":"Li","year":"2021","journal-title":"arXiv:2110.05679"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00064"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00502"},{"key":"ref26","article-title":"Micro-batch training with batch-channel normalization and weight standardization","author":"Qiao","year":"2019","journal-title":"arXiv:1903.10520"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1145\/3523273"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58558-7_29"},{"key":"ref29","article-title":"BYOL works even without batch statistics","author":"Richemond","year":"2020","journal-title":"arXiv:2010.10241"},{"key":"ref30","article-title":"DP-SGD vs PATE: Which has less disparate impact on model accuracy?","author":"Uniyal","year":"2021","journal-title":"arXiv:2106.12576"},{"key":"ref31","volume-title":"TensorFlow Privacy","year":"2024"},{"key":"ref32","volume-title":"TensorFlow Privacy (Membership Inference Attacks)","year":"2024"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00342"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01014"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00446"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1137\/0330046"},{"key":"ref37","article-title":"The trade-offs of private prediction","author":"van der Maaten","year":"2020","journal-title":"arXiv:2007.05089"},{"key":"ref38","article-title":"Augment your batch: Better training with larger batches","author":"Hoffer","year":"2019","journal-title":"arXiv:1901.09335"},{"key":"ref39","first-page":"1309","article-title":"Essentially no barriers in neural network energy landscape","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Draxler"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1145\/3314183.3323847"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00918"},{"key":"ref42","article-title":"R\u00e9nyi differential privacy of the sampled Gaussian mechanism","author":"Mironov","year":"2019","journal-title":"arXiv:1908.10530"},{"key":"ref43","article-title":"Knock knock, who\u2019s there? Membership inference on aggregate location data","author":"Pyrgelis","year":"2017","journal-title":"arXiv:1708.06145"},{"key":"ref44","first-page":"19964","article-title":"Batch normalization biases residual blocks towards the identity function in deep networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"33","author":"De"},{"key":"ref45","article-title":"ML-leaks: Model and data independent membership inference attacks and defenses on machine learning models","author":"Salem","year":"2018","journal-title":"arXiv:1806.01246"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/CSF.2018.00027"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1145\/2976749.2978355"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2021.3103191"},{"key":"ref49","article-title":"Differential privacy has bounded impact on fairness in classification","author":"Mangold","year":"2022","journal-title":"arXiv:2210.16242"},{"key":"ref50","article-title":"Learning multiple layers of features from tiny images","author":"Krizhevsky","year":"2009"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1109\/SP.2017.41"},{"key":"ref52","article-title":"Toward training at ImageNet scale with differential privacy","author":"Kurakin","year":"2022","journal-title":"arXiv:2201.12328"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1007\/11681878_14"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00934"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1201\/9781003278290-22"},{"key":"ref56","article-title":"Differentially private release of Israel\u2019s national registry of live births","author":"Hod","year":"2024","journal-title":"arXiv:2405.00267"},{"key":"ref57","first-page":"1","article-title":"Differential privacy has disparate impact on model accuracy","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"32","author":"Bagdasaryan"},{"key":"ref58","first-page":"6944","article-title":"Robin hood and Matthew effects: Differential privacy has disparate impact on synthetic data","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Ganev"},{"key":"ref59","article-title":"Google COVID-19 vaccination search insights: Anonymization process description","author":"Bavadekar","year":"2021","journal-title":"arXiv:2107.01179"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.1145\/3630106.3659006"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00894"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1007\/s13347-021-00478-z"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1016\/j.cose.2020.101951"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00794"},{"key":"ref65","first-page":"1","article-title":"A privacy-preserving deep learning approach for face recognition with edge computing","volume-title":"Proc. USENIX Workshop Hot Topics Edge Comput. (HotEdge)","author":"Mao"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.1109\/WACVW50321.2020.9096947"},{"key":"ref67","first-page":"15","article-title":"A review of face recognition against longitudinal child faces","volume-title":"Proc. BIOSIG","author":"Ricanek"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1109\/WACVW.2019.00023"},{"key":"ref69","doi-asserted-by":"publisher","DOI":"10.1109\/FG.2018.00020"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46487-9_6"},{"key":"ref71","article-title":"Labeled faces in the wild: A database for studying face recognition in unconstrained environments","volume-title":"Proc. Workshop Faces \u2018Real-Life\u2019 Images: Detection, Alignment, Recognit.","author":"Huang"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298803"},{"key":"ref73","article-title":"Learning face representation from scratch","author":"Yi","year":"2014","journal-title":"arXiv:1411.7923"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00078"},{"key":"ref75","article-title":"Inference attacks against face recognition model without classification layers","author":"Huang","year":"2024","journal-title":"arXiv:2401.13719"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1145\/2976749.2978318"},{"key":"ref77","volume-title":"Biometrics Large-Scale IT","year":"2024"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.29012\/jpc.722"},{"key":"ref79","article-title":"A database for studying crosspose face recognition in unconstrained environments","author":"Zheng","year":"2018"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW.2017.250"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.1109\/WACV.2016.7477558"},{"key":"ref82","article-title":"Cross-age LFW: A database for studying cross-age face recognition in unconstrained environments","author":"Zheng","year":"2017","journal-title":"arXiv:1708.08197"},{"key":"ref83","first-page":"1","article-title":"EfficientNet: Rethinking model scaling for convolutional neural networks","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Tan"},{"key":"ref84","volume-title":"A List of Real-World Uses of Differential Privacy","author":"Desfontaines","year":"2021"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01223"}],"container-title":["IEEE Access"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/6287639\/10820123\/10858162.pdf?arnumber=10858162","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,2,20]],"date-time":"2025-02-20T20:34:29Z","timestamp":1740083669000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10858162\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025]]},"references-count":85,"URL":"https:\/\/doi.org\/10.1109\/access.2025.3536784","relation":{},"ISSN":["2169-3536"],"issn-type":[{"value":"2169-3536","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025]]}}}