ARX

Free desktop software and Java library for anonymising structured personal data, with built-in checks of re-identification risk and of how much useful information remains.

What it is¶

ARX lets you load a table, mark which columns are direct identifiers, quasi-identifiers or sensitive attributes, and define generalisation hierarchies (for example, exact age to five-year bands). It then searches for a transformation that meets privacy models you choose, such as k-anonymity, l-diversity, t-closeness or differential privacy, while keeping as much data utility as possible. Its graphical interface shows the estimated re-identification risk before and after anonymisation, which helps you document why the output is safe to share. ARX grew out of medical informatics research at the Technical University of Munich and is now developed together with the Berlin Institute of Health at Charité.

In healthcare¶

ARX was built with biomedical data in mind and its developers present it as a tool for sharing research and clinical data. For a clinical AI project you can use it on tabular extracts, such as patient characteristics, diagnoses and lab results, before you share or publish them. It does not handle free text, DICOM headers or images, so you need other tools for those.

Where to use it in The ELSA Way¶

• Privacy and security › Pseudonymisation and anonymisation: ARX lets you apply and test formal anonymisation models on tabular patient data and measure the remaining re-identification risk, instead of assuming that removing names is enough.

References¶

• Prasser F, Eicher J, Spengler H, Bild R, Kuhn KA. Flexible data anonymization using ARX: Current status and challenges ahead. Software: Practice and Experience. 2020;50(7):1277-1304. Prasser et al. (2020)

Related tools¶

• Flower: An open-source Python framework for federated learning: you train one model across several sites while patient data stays where it is.