Privacy in Image Datasets: A Case Study on Pregnancy Ultrasounds
Not on the map. Why?
Abstract (arXiv)
The rise of generative models has led to increased use of large-scale datasets collected from the internet, often with minimal or no data curation. This raises concerns about the inclusion of sensitive or private information. In this work, we explore the presence of pregnancy ultrasound images, which contain sensitive personal information and are often shared online. Through a systematic examination of LAION-400M dataset using CLIP embedding similarity, we retrieve images containing pregnancy ultrasound and detect thousands of entities of private information such as names and locations. Our findings reveal that multiple images have high-risk information that could enable re-identification or impersonation. We conclude with recommended practices for dataset curation, data privacy, and ethical use of public image datasets.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 31 of its 79 references that are in the map. First posted on arXiv 6 Feb 2026.
Builds on
31 of 79 references.
References · 31
- 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
- 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
- 2014Microsoft COCO: Common Objects in Contextcited by 17,498
- 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
- 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
- 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
- 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
- 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
- 2018UMAP: Uniform Manifold Approximation and Projection for Dimension Reductioncited by 2,841
- 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,346
- 2020Extracting Training Data from Large Language Modelscited by 2,144
- 2021LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairscited by 1,544
19 more
- 2021Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Datacited by 961
- 2023Extracting Training Data from Diffusion Modelscited by 723
- 2023DataComp: In search of the next generation of multimodal datasetscited by 635
- 2016Semi-supervised Knowledge Transfer for Deep Learning from Private Training Datacited by 347
- 2021Multimodal datasets: misogyny, pornography, and malignant stereotypescited by 326
- 2024MACE: Mass Concept Erasure in Diffusion Modelscited by 298
- 2018Scalable Private Learning with PATEcited by 266
- 2020Large image datasets: A pyrrhic win for computer vision?cited by 237
- 2022A Self-Supervised Descriptor for Image Copy Detectioncited by 175
- 2020Privacy Preserving Face Recognition Utilizing Differential Privacycited by 30
- 2019P3SGD: Patient Privacy Preserving SGD for Regularizing Deep CNNs in Pathological Image Classificationcited by 14
- 2022Privacy-Preserving Action Recognition via Motion Difference Quantizationcited by 14
- 2024Unstable Unlearning: The Hidden Risk of Concept Resurgence in Diffusion Modelscited by 14
- 2019Image Privacy Prediction Using Deep Neural Networkscited by 13
- 2024Rethinking Artistic Copyright Infringements in the Era of Text-to-Image Generative Modelscited by 9
- 2024De-Identification of Medical Imaging Data: A Comprehensive Tool for Ensuring Patient Privacycited by 7
- 2025Neglected Risks: The Disturbing Reality of Children's Images in Datasets and the Urgent Call for Accountabilitycited by 6
- 2025A Common Pool of Privacy Problems: Legal and Technical Lessons from a Large-Scale Web-Scraped Machine Learning Datasetcited by 5
- 2022Scrape, Cut, Paste and Learn: Automated Dataset Generation Applied to Parcel Logisticscited by 4
Built on by
Not cited yet.