Fully-Adaptive Composition in Differential Privacy
Authors: Justin Whitehouse, Aaditya Ramdas, Ryan Rogers, Steven Wu
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | We show that, as long as a target privacy level is prespecified, one can obtain the same rate as advanced composition, including constants. We also construct families of privacy odometers that are not only tighter than the originals, but can be optimized for various target levels of privacy. Overall, we show that full adaptivity is not a cost but rather a feature of differential privacy. Our key insight is to view adaptive privacy composition as depending not on the number of algorithms being composed, but rather on the sums of squares of privacy parameters, P m n ϵ2 m. This shift to looking at intrinsic time allows us to apply recent advances in time-uniform concentration (Howard et al., 2020; 2021) to privacy loss martingales. |
| Researcher Affiliation | Collaboration | 1Carnegie Mellon University 2Linked In. Correspondence to: Justin Whitehouse <jwhiteho@andrew.cmu.edu>. |
| Pseudocode | No | The paper does not contain any sections labeled 'Algorithm' or 'Pseudocode', nor does it present structured steps of a method in a code-like format. |
| Open Source Code | No | The paper does not contain any statements about open-sourcing code, nor does it provide links to a code repository for the methodology described. |
| Open Datasets | No | The paper is theoretical and focuses on mathematical derivations and analysis of privacy composition, rather than empirical evaluation on datasets. Therefore, no information about publicly available datasets is provided. |
| Dataset Splits | No | The paper is theoretical and does not involve empirical evaluation on datasets, thus there are no training, validation, or test splits mentioned. |
| Hardware Specification | No | The paper is theoretical and presents mathematical derivations and analysis, with no described experiments that would require specific hardware. No hardware specifications are mentioned. |
| Software Dependencies | No | The paper is purely theoretical, focusing on mathematical concepts and proofs. It does not describe any implemented software or empirical evaluations that would require listing software dependencies with version numbers. |
| Experiment Setup | No | The paper is theoretical and does not include any experimental evaluations or computational simulations that would require details on experimental setup or hyperparameters. |