Optimal conversion from Rényi Differential Privacy to f f -Differential Privacy
Anneliese Riess, Juan Felipe Gomez, Flavio du Pin Calmon, Julia Anne Schnabel, Georgios Kaissis · 2026
Read Optimal conversion from Rényi Differential Privacy to f f -Differential Privacy by Anneliese Riess, Juan Felipe Gomez, Flavio du Pin Calmon, Julia Anne Schnabel, Georgios Kaissis (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-2602-04562.txt.
Description
Anneliese Riess and 4 co-authors settle a conjecture about privacy accounting in a 2026 paper. Differential privacy limits how much one person's data can affect an output. Researchers often track privacy using Rényi differential privacy, then convert it to a hypothesis-testing form called f-differential privacy.
The authors prove that one known conversion rule is optimal among all possible rules. The rule takes the pointwise maximum over bounds from each single privacy order. Their proof uses the geometry of the privacy region, whose boundary is set by simple Bernoulli mechanisms. No black-box conversion can do uniformly better using only Rényi guarantees. The result marks a fundamental limit for privacy bookkeeping.
Catalog record
| CREATOR | Anneliese Riess, Juan Felipe Gomez, Flavio du Pin Calmon, Julia Anne Schnabel, Georgios Kaissis |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-04562