Natural Privacy Filters Are Not Always Free: A Characterization of Free Natural Filters
2026-02-17 • Cryptography and Security
Cryptography and SecurityData Structures and Algorithms
AI summaryⓘ
The authors study a method called natural privacy filters, which help combine different privacy-protecting tools that adapt their settings as they go. Unlike earlier methods that looked at simpler privacy measures, natural filters consider the full privacy details of each tool, which could allow better performance without losing privacy. However, the authors found that these natural filters don't always work without cost; only certain well-structured sets of privacy tools can use these filters without additional privacy loss. This shows that natural privacy filters have limitations depending on the types of privacy mechanisms used.
differential privacyprivacy filtersprivacy compositionRényi differential privacyGaussian differential privacyadaptive mechanismsprivacy budgetprivacy profilewell-ordered families
Authors
Matthew Regehr, Bingshan Hu, Ethan Leeman, Pasin Manurangsi, Pierre Tholoniat, Mathias Lécuyer
Abstract
We study natural privacy filters, which enable the exact composition of differentially private (DP) mechanisms with adaptively chosen privacy characteristics. Earlier privacy filters consider only simple privacy parameters such as Rényi-DP or Gaussian DP parameters. Natural filters account for the entire privacy profile of every query, promising greater utility for a given privacy budget. We show that, contrary to other forms of DP, natural privacy filters are not free in general. Indeed, we show that only families of privacy mechanisms that are well-ordered when composed admit free natural privacy filters.