Research map / Federated learning and privacy / Differential privacy
Unified Enhancement of Privacy Bounds for Mixture Mechanisms via $f$-Differential Privacy
Chendi Wang, Buxin Su, Jiayuan Ye, Reza Shokri, Weijie J Su
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use f-DP joint concavity-based bound instead of prior (eps,delta)-DP analysis (Thm 3.2) | shuffled mechanism · eps=0.5, n=10000, eps0=4.444 | delta: 0.9494 -> $3 \times 10^{-6}$ | Improved |
| same, at eps=0.7 | shuffled mechanism · eps=0.7 | delta: 0.1038 -> $4 \times 10^{-9}$ | Improved |
| same, at eps=1.0 | shuffled mechanism · eps=1.0 | delta: $8 \times 10^{-5}$ -> $2 \times 10^{-14}$ | Improved |
| use f-DP bound instead of prior numerical upper bound | shuffled mechanism · delta=5e-5 | epsilon: 1.014 -> $0.4$ | Improved |
| use f-DP bound instead of prior numerical upper bound | shuffled mechanism · delta=3e-6 | epsilon: 1.085 -> 0.5 | Improved |
| drop the eps0 <= log(n/(8 log(2/delta)) - 1) assumption | shuffled mechanism bound · delta<=1e-7 | bound validity | Improved |
| our eps bound vs empirical lower bound | f-DP bound · delta=9e-11 | epsilon: 0.8 vs 0.758 | Improved |
| advanced joint concavity instead of joint concavity (ablation) | trade-off function bound · w=1/3, eps=0.5 | delta: 0.0020 -> $1.5 \times 10^{-6}$ | Improved |
| random initialization in one-step DP-GD vs c-GDP ignoring initialization randomness | DP-GD · one-step DP-GD | privacy trade-off | Improved |
Closest papers by meaning
- Shifted Interpolation for Differential Privacy (ICML 2024 · similarity 0.91)
- Beyond the Calibration Point: Mechanism Comparison in Differential Privacy (ICML 2024 · similarity 0.90)
- Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement (NeurIPS 2024 · similarity 0.89)
- Mitigating the Privacy–Utility Trade-off in Decentralized Federated Learning via f-Differential Privacy (NeurIPS 2025 · similarity 0.89)
- Sampling-Free Privacy Accounting for Matrix Mechanisms under Random Allocation (NeurIPS 2026 · similarity 0.89)
- 𝑓-Differential Privacy Filters: Validity and Approximate Solutions (NeurIPS 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
