On Choosing the μ Parameter in Gaussian Differential Privacy
Not on the map. Why?
Abstract (arXiv)
Recent work argues for using Gaussian differential privacy (GDP) to report the privacy guarantees in privacy-preserving machine learning. We provide principled mappings from pure-DP varepsilon to GDP μ by matching the worst-case success of a strong-adversary membership inference attack in terms of three metrics: multiplicative advantage at fixed FPR, precision at fixed recall, and the standard privacy profile. We tabulate μ values across a useful range of parameters and recommend μ ≈ varepsilon/5 as a conservative general-purpose conversion.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 5 of its 10 references that are in the map. First posted on arXiv 8 Jun 2026.
Builds on
5 of 10 references.
References · 5
- 2013The Composition Theorem for Differential Privacycited by 297
- 2018Privacy Amplification by Subsampling: Tight Analyses via Couplings and Divergencescited by 242
- 2016Differential Privacy as a Mutual Information Constraintcited by 45
- 2019SoK: Differential Privaciescited by 34
- 2024Attack-Aware Noise Calibration for Differential Privacycited by 8
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2022Generalised Likelihood Ratio Testing Adversaries through the Differential Privacy Lens
- 2024Evaluations of Machine Learning Privacy Defenses are Misleading
- 2021Generalization Techniques Empirically Outperform Differential Privacy against Membership Inference
- 2022Provable Membership Inference Privacy
- 2024Calibrating Practical Privacy Risks for Differentially Private Machine Learning
- 2023Gaussian Membership Inference Privacy
- 2023Epsilon*: Privacy Metric for Machine Learning Models
- 2025Empirical Calibration and Metric Differential Privacy in Language Models