Research map / Federated learning and privacy / Differential privacy
Privacy Auditing with One (1) Training Run
Thomas Steinke, Milad Nasr, Matthew Jagielski
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 |
|---|---|---|---|
| increase number of auditing examples m (ablation) | one-run audit · white-box CIFAR-10 | epsilon lower bound | Improved |
| add more non-auditing training examples (n-m) (ablation) | one-run audit · white-box CIFAR-10 | epsilon lower bound | No clear effect |
| increase number of DP-SGD iterations at equal privacy (ablation) | one-run audit · white-box CIFAR-10 | epsilon lower bound | No clear effect |
| increase number of auditing examples m (ablation) | one-run audit · black-box input-space | epsilon lower bound | Mixed |
| add superfluous training data (ablation) | one-run audit · black-box input-space | epsilon lower bound | Hurt |
| insert auditing canaries in DP-SGD training | DP-SGD · CIFAR-10 epsilon=8 | test accuracy: 78 -> 76 | Hurt |
| increase number of guesses r in idealized randomized response (ablation) | one-run audit · idealized epsilon=4 | epsilon lower bound | Improved |
| increase number of guesses r in idealized Gaussian setting (ablation) | one-run audit · idealized Gaussian | epsilon lower bound | Hurt |
Closest papers by meaning
- How Well Can Differential Privacy Be Audited in One Run? (NeurIPS 2025 · similarity 0.93)
- Auditing Private Prediction (ICML 2024 · similarity 0.91)
- Nearly Tight Black-Box Auditing of Differentially Private Machine Learning (NeurIPS 2024 · similarity 0.90)
- Auditing $f$-differential privacy in one run (ICML 2025 · similarity 0.90)
- Tighter Privacy Auditing of DP-SGD in the Hidden State Threat Model (ICLR 2025 · similarity 0.89)
- One-shot Empirical Privacy Estimation for Federated Learning (ICLR 2024 · 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.
