{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T19:08:38Z","timestamp":1784833718717,"version":"3.55.0"},"reference-count":40,"publisher":"Public Library of Science (PLoS)","issue":"7","license":[{"start":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T00:00:00Z","timestamp":1784764800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100002347","name":"Bundesministerium f\u00fcr Bildung und Forschung","doi-asserted-by":"publisher","award":["KIS60FAC004"],"award-info":[{"award-number":["KIS60FAC004"]}],"id":[{"id":"10.13039\/501100002347","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.plosdigitalhealth.org"],"crossmark-restriction":false},"short-container-title":["PLOS Digit Health"],"abstract":"<jats:p>Advances in realistic image synthesis have led to several downstream applications in healthcare, particularly the anonymisation of biometric images. However, generative models pose significant privacy risks when synthetic outputs closely resemble the real, personal images required for training. To address these challenges, we propose a two-stage framework for generating privacy-preserving synthetic eye images based on a dataset of real photographs. Our system consists of a fine-tuned text-to-image synthesiser based on Stable Diffusion (v1-5), followed by a privacy risk scorer termed the \u201cCone of Privacy\u201d (CoP). Inspired by the DreamBooth method, the synthesiser incorporates information about the original dataset to generate similar realistic yet diverse images. To mitigate identity leakage, the CoP score measures the identifiability of synthetic images with respect to the training images based on embeddings from a pre-trained vision transformer. In evaluations on a dataset of 2,000 images from 704 subjects, our approach successfully generated realistic images while filtering high-risk samples, achieving a privacy threat detection rate above 97%. The Fr\u00e9chet Inception Distance (FID) between the original and the fully synthetic datasets was 57.1, with privacy-violating synthetic images scoring 51.3 and privacy-preserving images scoring 82.6, demonstrating a trade-off between realism and privacy protection. These results suggest the framework is a promising solution for anonymising biometric eye images in clinical and research applications.<\/jats:p>","DOI":"10.1371\/journal.pdig.0001487","type":"journal-article","created":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T18:54:16Z","timestamp":1784832856000},"page":"e0001487","update-policy":"https:\/\/doi.org\/10.1371\/journal.pdig.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["Ensuring data protection for eye images by combining fine-tuned image synthesis and anonymity assessment"],"prefix":"10.1371","volume":"5","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-3108-6495","authenticated-orcid":true,"given":"Sebastian","family":"Uschmann","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7301-1566","authenticated-orcid":true,"given":"Cord","family":"Spreckelsen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sven","family":"Festag","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2026,7,23]]},"reference":[{"key":"pdig.0001487.ref001","doi-asserted-by":"crossref","first-page":"102846","DOI":"10.1016\/j.media.2023.102846","article-title":"Diffusion models in medical imaging: a comprehensive survey","volume":"88","author":"A Kazerouni","year":"2023","journal-title":"Med Image Anal"},{"key":"pdig.0001487.ref002","doi-asserted-by":"crossref","first-page":"103046","DOI":"10.1016\/j.media.2023.103046","article-title":"Deep learning based synthesis of MRI, CT and PET: review and analysis","volume":"92","author":"S Dayarathna","year":"2024","journal-title":"Med Image Anal"},{"issue":"10","key":"pdig.0001487.ref003","doi-asserted-by":"crossref","first-page":"1430","DOI":"10.1136\/bjo-2023-324923","article-title":"Using artificial intelligence to improve human performance: efficient retinal disease detection training with synthetic images","volume":"108","author":"H Tabuchi","year":"2024","journal-title":"Br J Ophthalmol"},{"key":"pdig.0001487.ref004","doi-asserted-by":"crossref","first-page":"105042","DOI":"10.1016\/j.jdent.2024.105042","article-title":"Synthetic, non-person related panoramic radiographs created by generative adversarial networks in research, clinical, and teaching applications","volume":"146","author":"R Schoenhof","year":"2024","journal-title":"J Dent"},{"issue":"8","key":"pdig.0001487.ref005","doi-asserted-by":"crossref","first-page":"1257","DOI":"10.1007\/s11548-020-02185-0","article-title":"The effects of different levels of realism on the training of CNNs with only synthetic images for the semantic segmentation of robotic instruments in a head phantom","volume":"15","author":"SA Heredia Perez","year":"2020","journal-title":"Int J Comput Assist Radiol Surg"},{"key":"pdig.0001487.ref006","doi-asserted-by":"crossref","unstructured":"Moghadam PA, Van Dalen S, Martin KC, Lennerz J, Yip S, Farahani H, et al. A Morphology Focused Diffusion Probabilistic Model for Synthesis of Histopathology Images. In: Proceedings of the IEEE\/CVF Winter Conference on Applications of Computer Vision (WACV). 2023. pp. 2000\u2013-9. https:\/\/doi.org\/10.1109\/WACV56688.2023.00204","DOI":"10.1109\/WACV56688.2023.00204"},{"key":"pdig.0001487.ref007","doi-asserted-by":"crossref","first-page":"117","DOI":"10.1007\/978-3-031-18576-2_12","article-title":"Brain Imaging Generation with Latent Diffusion Models.","volume-title":"Deep Generative Models","author":"WHL Pinaya","year":"2022"},{"key":"pdig.0001487.ref008","first-page":"71","article-title":"Synthetic data and GDPR compliance: how artificial intelligence might resolve the privacy-utility tradeoff","volume":"28","author":"M Cairo","year":"2023","journal-title":"J Tech L Pol\u2019y"},{"key":"pdig.0001487.ref009","doi-asserted-by":"crossref","first-page":"106","DOI":"10.1007\/978-3-030-76663-4_6","article-title":"Towards Improving Privacy of Synthetic DataSets.","volume-title":"Privacy Technologies and Policy","author":"A Kuppa","year":"2021"},{"key":"pdig.0001487.ref010","first-page":"5253","article-title":"Extracting training data from diffusion models.","volume-title":"Proceedings of the 32nd USENIX Conference on Security Symposium. SEC \u201923","author":"N Carlini","year":"2023"},{"key":"pdig.0001487.ref011","unstructured":"European Parliament and Council of the European Union. Regulation (EU) 2016\/679 of the European Parliament and of the Council. 2016. Available from: https:\/\/eur-lex.europa.eu\/eli\/reg\/2016\/679\/oj"},{"key":"pdig.0001487.ref012","unstructured":"Illinois General Assembly. Biometric Information Privacy Act (BIPA). 740 ILCS 14\/1\u201314\/99. 2008. Available from: https:\/\/www.ilga.gov\/legislation\/ilcs\/ilcs5.asp?ActID=3004"},{"key":"pdig.0001487.ref013","unstructured":"U S Congress. Health Insurance Portability and Accountability Act (HIPAA) Public Law 104-191, 110 Stat. 1936\u20132103. 1996. Available from: https:\/\/www.govinfo.gov\/content\/pkg\/PLAW-104publ191\/pdf\/PLAW-104publ191.pdf"},{"issue":"3","key":"pdig.0001487.ref014","doi-asserted-by":"crossref","first-page":"69","DOI":"10.3390\/jimaging9030069","article-title":"GANs for medical image synthesis: an empirical study","volume":"9","author":"Y Skandarani","year":"2023","journal-title":"J Imaging"},{"issue":"1","key":"pdig.0001487.ref015","doi-asserted-by":"crossref","first-page":"167","DOI":"10.3390\/s25010167","article-title":"WGAN-GP for synthetic retinal image generation: enhancing sensor-based medical imaging for classification models","volume":"25","author":"H Anaya-S\u00e1nchez","year":"2024","journal-title":"Sensors (Basel)"},{"key":"pdig.0001487.ref016","first-page":"395","article-title":"Diffusion Models for Sign Language Video Anonymization.","volume-title":"Proceedings of the LREC-COLING 2024 11th Workshop on the Representation and Processing of Sign Languages: Evaluation of Sign Language Resources","author":"Z Xia","year":"2024"},{"key":"pdig.0001487.ref017","doi-asserted-by":"crossref","first-page":"3","DOI":"10.1007\/978-3-031-53767-7_1","article-title":"Privacy distillation: Reducing re-identification risk of diffusion models","volume-title":"Deep generative models","author":"V Fernandez","year":"2024"},{"key":"pdig.0001487.ref018","doi-asserted-by":"crossref","unstructured":"Packh\u00e4user K, Folle L, Thamm F, Maier A. Generation of Anonymous Chest Radiographs Using Latent Diffusion Models for Training Thoracic Abnormality Classification Systems. In: 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI). 2023. pp. 1\u20135. https:\/\/doi.org\/10.1109\/ISBI53787.2023.10230346","DOI":"10.1109\/ISBI53787.2023.10230346"},{"key":"pdig.0001487.ref019","first-page":"2672","article-title":"Generative Adversarial Nets.","volume-title":"Proceedings of the 28th International Conference on Neural Information Processing Systems - Volume 2. NIPS\u201914","author":"IJ Goodfellow","year":"2014"},{"key":"pdig.0001487.ref020","doi-asserted-by":"crossref","first-page":"10674","DOI":"10.1109\/CVPR52688.2022.01042","article-title":"High-Resolution Image Synthesis with Latent Diffusion Models.","volume-title":"2022 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)","author":"R Rombach","year":"2022"},{"key":"pdig.0001487.ref021","doi-asserted-by":"crossref","unstructured":"Kumari N, Zhang B, Wang SY, Shechtman E, Zhang R, Zhu JY. Ablating Concepts in Text-to-Image Diffusion Models. In: 2023 IEEE\/CVF International Conference on Computer Vision (ICCV). 2023. pp. 22634\u201345. https:\/\/doi.org\/10.1109\/ICCV51070.2023.02074","DOI":"10.1109\/ICCV51070.2023.02074"},{"key":"pdig.0001487.ref022","unstructured":"Webster R. A Reproducible Extraction of Training Images from Diffusion Models. arXiv; 2023. http:\/\/arxiv.org\/abs\/2305.08694 https:\/\/doi.org\/10.48550\/arXiv.2305.08694"},{"key":"pdig.0001487.ref023","doi-asserted-by":"crossref","first-page":"6048","DOI":"10.1109\/CVPR52729.2023.00586","article-title":"Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models.","volume-title":"2023 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR)","author":"G Somepalli","year":"2023"},{"key":"pdig.0001487.ref024","unstructured":"Usynin D, Knolle M, Kaissis G. Memorisation in Machine Learning: A Survey of Results. Transact Mach Learn Res. 2024. Available from: https:\/\/openreview.net\/forum?id=HVWODwbrFK"},{"key":"pdig.0001487.ref025","unstructured":"Wen Y, Liu Y, Chen C, Lyu L. Detecting, Explaining, and Mitigating Memorization in Diffusion Models. In: The Twelfth International Conference on Learning Representations. 2024. Available from: https:\/\/openreview.net\/forum?id=84n3UwkH7b"},{"key":"pdig.0001487.ref026","first-page":"47783","volume-title":"Proceedings of the 37th International Conference on Neural Information Processing Systems","author":"G Somepalli","year":"2023"},{"key":"pdig.0001487.ref027","doi-asserted-by":"crossref","first-page":"340","DOI":"10.1007\/978-3-031-72980-5_20","article-title":"Unveiling and Mitigating Memorization in Text-to-Image Diffusion Models Through Cross Attention.","volume-title":"Computer Vision \u2013 ECCV 2024","author":"J Ren","year":"2024"},{"issue":"8","key":"pdig.0001487.ref028","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3721479","article-title":"Attacks and defenses for generative diffusion models: a comprehensive survey","volume":"57","author":"VT Truong","year":"2025","journal-title":"ACM Comput Surv"},{"key":"pdig.0001487.ref029","doi-asserted-by":"crossref","unstructured":"Pang Y, Wang T. Black-box membership inference attacks against fine-tuned diffusion models. 2024. https:\/\/doi.org\/10.14722\/ndss.2025.23324","DOI":"10.14722\/ndss.2025.230324"},{"key":"pdig.0001487.ref030","doi-asserted-by":"crossref","unstructured":"Ruiz N, Li Y, Jampani V, Pritch Y, Rubinstein M, Aberman K. DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation. In: Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2023. pp. 22500\u201310. https:\/\/doi.org\/10.1109\/CVPR52729.2023.02155","DOI":"10.1109\/CVPR52729.2023.02155"},{"key":"pdig.0001487.ref031","unstructured":"Lu Y, Singhal S, Strub F, Pietquin O, Courville A. Countering language drift with seeded iterated learning. In: Proceedings of the 37th International Conference on Machine Learning. ICML\u201920. JMLR.org; 2020. pp. 6437\u20136447. https:\/\/doi.org\/10.5555\/3524938.3525535"},{"key":"pdig.0001487.ref032","unstructured":"Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai X, Unterthiner T, et al. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. arXiv; 2021. https:\/\/doi.org\/10.48550\/arXiv.2010.11929"},{"key":"pdig.0001487.ref033","doi-asserted-by":"crossref","first-page":"248","DOI":"10.1109\/CVPR.2009.5206848","article-title":"Imagenet: A large-scale hierarchical image database.","volume-title":"2009 IEEE conference on computer vision and pattern recognition","author":"J Deng","year":"2009"},{"key":"pdig.0001487.ref034","unstructured":"Oquab M, Darcet T, Moutakanni T, Vo HV, Szafraniec M, Khalidov V, et al. DINOv2: Learning Robust Visual Features without Supervision. Transactions on Machine Learning Research. Featured Certification. 2024. Available from: https:\/\/openreview.net\/forum?id=a68SUt6zFt"},{"key":"pdig.0001487.ref035","doi-asserted-by":"crossref","first-page":"1534","DOI":"10.1109\/WACV51458.2022.00160","article-title":"Visualizing Paired Image Similarity in Transformer Networks.","volume-title":"2022 IEEE\/CVF Winter Conference on Applications of Computer Vision (WACV)","author":"S Black","year":"2022"},{"issue":"1","key":"pdig.0001487.ref036","doi-asserted-by":"crossref","first-page":"3","DOI":"10.1145\/1217299.1217302","article-title":"L-diversity: Privacy beyond k-anonymity","volume":"1","author":"A Machanavajjhala","year":"2007","journal-title":"ACM Trans Knowl Discov Data"},{"key":"pdig.0001487.ref037","article-title":"Conjunctivitis","volume-title":"StatPearls","author":"MF Hashmi","year":"2025"},{"key":"pdig.0001487.ref038","article-title":"Iris Super Resolution Dataset","author":"S Aryanmehr","year":"2019","journal-title":"IEEE Dataport"},{"key":"pdig.0001487.ref039","first-page":"6629","article-title":"GANs trained by a two time-scale update rule converge to a local nash equilibrium.","volume-title":"Proceedings of the 31st International Conference on Neural Information Processing Systems","author":"M Heusel","year":"2017"},{"key":"pdig.0001487.ref040","unstructured":"Court of Justice of the European Union. Judgment of 4 September 2025 \u2013 Case C-413\/23 P; ECLI:EU:C:2025:645. 2025. Available from: https:\/\/eur-lex.europa.eu\/legal-content\/EN\/TXT\/?uri=celex:62023CJ0413"}],"container-title":["PLOS Digital Health"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pdig.0001487","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T18:54:22Z","timestamp":1784832862000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pdig.0001487"}},"subtitle":[],"editor":[{"given":"Sulaf","family":"Assi","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,7,23]]},"references-count":40,"journal-issue":{"issue":"7","published-online":{"date-parts":[[2026,7,23]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pdig.0001487","relation":{},"ISSN":["2767-3170"],"issn-type":[{"value":"2767-3170","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,23]]}}}