Research map / Federated learning and privacy / Differential privacy
The Price of Differential Privacy under Continual Observation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Require continual release rather than one batch output | Differentially private maxsum · maxsum, approximate DP | worst-case error: The continual-release lower bound is $\tilde \Omega\Big(\min\Big\{\sqrt[3]{\frac T {\varepsilon^2}},\frac{\sqrt{d}}\vare | Hurt |
| Require continual release rather than one batch output | Differentially private maxsum · maxsum, pure DP | worst-case error: $\tilde \Omega\Big(\min\Big\{\sqrt{\frac T \varepsilon},\frac{d}\varepsilon,T\Big\}\Big)$ versus $O(\frac 1\varepsilon)$ | Hurt |
| Require continual release rather than one batch output | Differentially private selection · selection, approximate DP | worst-case error: The continual-release lower bound is $\tilde \Omega\Big(\min\Big\{\sqrt[3]{\frac {T\log^2 d} {\varepsilon^2}},\frac{\sqr | Hurt |
| Require continual release rather than one batch output | Differentially private selection · selection, pure DP | worst-case error: $\tilde \Omega\Big(\min\Big\{\sqrt{\frac {T\log d} {\varepsilon}},\frac{d}\varepsilon,T\Big\}\Big)$ versus $O\paren{\fra | Hurt |
| Use the stated continual-release mechanisms instead of an unrestricted mechanism | maxsum and selection · maxsum and selection, continual release | worst-case error: The upper and lower bounds differ only in the $\text{polylog}(T)$ terms | No clear effect |
Closest papers by meaning
- Differentially Private Continual Release with Relative Error (ICML 2026 · similarity 0.93)
- Private Continual Counting of Unbounded Streams (NeurIPS 2025 · similarity 0.88)
- Continual Counting with Gradual Privacy Expiration (NeurIPS 2024 · similarity 0.88)
- Privacy Amplification Persists under Unlimited Synthetic Data Release (NeurIPS 2026 · similarity 0.88)
- Privacy Guarantees in Posterior Sampling under Contamination (NeurIPS 2026 · similarity 0.87)
- Exactly Minimax-Optimal Locally Differentially Private Sampling (NeurIPS 2024 · similarity 0.87)
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.