The DOI matches its Crossref record.
Composition in Differential Privacy for General Granularity Notions
Open accessAbstract
The composition theorems of differential privacy (DP) allow data curators to combine different algorithms to obtain a new algorithm that continues to satisfy DP. However, new granularity notions (i.e., neigh-borhood definitions), data domains, and composition settings have appeared in the literature that the classical composition theorems do not cover. For instance, the original parallel composition theorem does not translate well to general granularity notions. This complicates the opportunity of composing DP mechanisms in new settings and obtaining accurate estimates of the incurred privacy loss after composition. To overcome these limitations, we study the composability of DP in a general framework and for any kind of data domain or neighborhood definition. We give a general composition theorem in both independent and adaptive versions and we provide analogous composition results for approximate, zero-concentrated, and Gaussian DP. Besides, we study the hypothesis needed to obtain the best composition bounds. Our theorems cover both parallel and sequential composition settings. Importantly, they also cover every setting in between, allowing us to compute the final privacy loss of a composition with greatly improved accuracy.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Advanced composition frameworks, such as Rényi DP or zero-concentrated DP [39], offer tighter privacy bounds for iterative workflows, enabling longer training without budget exhaustion.”
From Privacy-preserving Machine Learning in Internet of Vehicle Applications: Fundamentals, Recent Advances, and Future Direction (2025) · cites this paper as [39]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
No references could be resolved yet.
Research works printed but not linked: 23
- 10.1007/978-3-662-53641-4_24 · DOI
- 10.29012/jpc.784 · DOI
- 10.1007/978-3-642-39077-7_5 · DOI
- Desfontaines, 2020, Ted is writing things (personal blog)
- 10.2478/popets-2020-0028 · DOI
- Dong, 2022, Gaussian Dif-ferential Privacy, J. Royal Stat. Soc. Ser. B, 84.1, 3, 10.1111/rssb.12454 · DOI
- 10.1007/11787006_1 · DOI
- 10.1007/11761679_29 · DOI · Paper page
- Dwork, 2014, The Algorithmic Foun-dations of Differential Privacy, Found. Trends Theor. Comput. Sci., 9.3-4, 211
- 10.5220/0011885000003405 · DOI
- 10.1109/ICDM.2009.11 · DOI
- 10.1145/1989323.1989345 · DOI
- Kifer, 2020, Guidelines for Implementing and Auditing Differentially Private Systems
- 10.2307/2533531 · DOI
- Levy, 2002, Basic Set Theory
- 10.2200/S00735ED1V01Y201609SPT018 · DOI
- 10.1145/1559845.1559850 · DOI
- Meiser, 2018, Approximate and Probabilistic Differential Privacy Definitions
- 10.1109/CSF.2017.11 · DOI · Paper page
- Soria-Comas, 2016, Big Data Privacy: Challenges to Privacy Principles and Mod-els, Data Sci. Eng. 1.1, 21, 10.1007/s41019-015-0001-x · DOI
- 10.1007/3-540-45523-X_17 · DOI
- van Erven, 2014, Renyi Divergence and Kullback-Leibler Divergence, IEEE Trans. Inform. Theory, 60.7, 3797, 10.1109/TIT.2014.2320500 · DOI
- See the proofs and further remarks in the appendix of the long version of this paper
Cited by
Sorted by citations
Islam, N., & Zulkernine, M. (2025) · arXiv · doi:10.48550/arxiv.2503.01089 · Cited by 9
“It aggregates data based on class distribution while keeping user data private on devices.”
Metadata
- Authors
- Guerra-Balboa, P., Miranda-Pascual, À., Parra-Arnau, J., & Strufe, T.
- Affiliations
- Karlsruhe Institute of Technology (DE)
- Universitat Politècnica de Catalunya,Spain
- Venue
- 2024 IEEE 37th Computer Security Foundations Symposium (CSF)
- Publisher
- IEEE
- Published
- 2024-07-08
- Pages
- 680-696
- ISSN
- 1063-6900 (ISSN-L)
- OpenAlex
- W4402674271
- Cited by
- 1
- References
- 0
- Citations by year
- 2026: 1
- FWCI
- 0.30
- Type
- conference paper
- Language
- en
- Open access
- green
- hdl.handle.net
- Locations
- IEEE Computer Security Foundations Workshop (CSFW) (publishedVersion)
- UPCommons institutional repository (Universitat Politècnica de Catalunya) (submittedVersion, cc-by-nc-nd)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- Physical Sciences › Computer Science › Information Systems › Cloud Data Security Solutions
- Physical Sciences › Computer Science › Computer Networks and Communications › Cryptography and Data Security
- Keywords
- differential privacy, composition theorem, privacy loss, neighborhood definition, data domains, parallel composition, sequential composition, approximate differential privacy, Gaussian differential privacy, privacy mechanisms
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
