Beyond Theoretical Bounds: Empirical Privacy Loss Calibration for Text Rewriting Under Local Differential Privacy
Weijun Li, Arnaud Grivet Sébert, Qiongkai Xu, Annabelle McIver, Mark Dras · 2026
Read Beyond Theoretical Bounds: Empirical Privacy Loss Calibration for Text Rewriting Under Local Differential Privacy by Weijun Li, Arnaud Grivet Sébert, Qiongkai Xu, Annabelle McIver, Mark Dras (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-2603-22968.txt.
Description
Li and co-authors propose a way to calibrate privacy loss for text rewriting systems. Such systems hide text under local differential privacy before release. The formal privacy value is hard to interpret and compare across tools. Their method, TeDA, frames calibration as hypothesis testing. It audits how distinguishable privatized texts remain, in both surface and embedding space. Applied to several tools, it shows equal nominal values can mean very different real exposure. The approach gives a shared scale for privacy-utility trade-offs.
Catalog record
| CREATOR | Weijun Li, Arnaud Grivet Sébert, Qiongkai Xu, Annabelle McIver, Mark Dras |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 65094 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.22968