Privacy in Image Datasets: A Case Study on Pregnancy Ultrasounds
Rawisara Lohanimit, Yankun Wu, Amelia Katirai, Yuta Nakashima, Noa Garcia · 2026
Read Privacy in Image Datasets: A Case Study on Pregnancy Ultrasounds by Rawisara Lohanimit, Yankun Wu, Amelia Katirai, Yuta Nakashima, Noa Garcia (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-07149.txt.
Description
Large image data sets scraped from the web can carry private material nobody checked for. Rawisara Lohanimit and four co-authors hunt for pregnancy ultrasound images in LAION-400M, a dataset widely used to train generative models, using image-embedding similarity to find them.
They recover ultrasound images and detect thousands of pieces of private information, such as names and places, some sharp enough to allow re-identification or impersonation. The paper closes with useful advice on curating data sets, protecting privacy, and using public image collections ethically. It is a concrete warning, grounded in one sensitive case, for everyone building on scraped data.
Catalog record
| CREATOR | Rawisara Lohanimit, Yankun Wu, Amelia Katirai, Yuta Nakashima, Noa Garcia |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-07149