On the Impact of Noise in Differentially Private Text Rewriting
Stephen Meisenbacher, Maulik Chevli, Florian Matthes · 2025
Read On the Impact of Noise in Differentially Private Text Rewriting by Stephen Meisenbacher, Maulik Chevli, Florian Matthes (2025) 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-2501-19022.txt.
Description
Rewriting private text with differential privacy means injecting noise, and noise eats meaning. Stephen Meisenbacher, Maulik Chevli, and Florian Matthes measure that tax in 2025 with a new sentence-infilling privatizer, comparing guaranteed-private rewriting against methods that promise nothing formal.
Across their tests, the non-private techniques kept text more useful and reached a decent informal balance, yet none could beat the formal methods at actually defeating privacy attacks. The size of the privacy parameter steered the damage, confirming noise as the main culprit in today's utility loss. The authors use the results to argue the field should be honest about what the guarantee costs and what shortcuts give up.
Catalog record
| CREATOR | Stephen Meisenbacher, Maulik Chevli, Florian Matthes |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-19022