Research map / Federated learning and privacy / Differential privacy
Have it your way: Individualized Privacy Assignment for DP-SGD
Franziska Boenisch, Christopher Mühl, Adam Dziedzic, Roy Rinberg, Nicolas Papernot
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 |
|---|---|---|---|
| Individualized sample rates (Sample) instead of uniform DP-SGD | DP-SGD · MNIST | test accuracy: 96.75 -> 97.81 | Improved |
| Individualized sample rates instead of uniform DP-SGD | DP-SGD · SVHN | test accuracy: 83.26 -> 85.89 | Improved |
| Individualized sample rates instead of uniform DP-SGD | DP-SGD · CIFAR10 | test accuracy: 52.77 -> 57.86 | Improved |
| Individualized noise and clipping (Scale) instead of uniform DP-SGD | DP-SGD · CIFAR10 | test accuracy: 52.77 -> 58.03 | Improved |
| Privacy group split 54%-37%-9% instead of 34%-43%-23% | IDP-SGD Sample · CIFAR10 | test accuracy: 57.86 -> 56.83 | Hurt |
| Sample vs Scale individualization | IDP-SGD · MNIST | test accuracy: 97.81 vs 97.78 | No clear effect |
| IDP-SGD on BERT fine-tuning SNLI | DP-SGD · SNLI | accuracy: 75.91 -> 76.5 | Improved |
| IDP-SGD on ResNet18 CIFAR10 | DP-SGD · CIFAR10 | accuracy: 47.52 -> 48.77 | Improved |
| IDP-SGD on embedding model | DP-SGD · IMDB | accuracy: 72.69 -> 73.34 | Improved |
| IDP-SGD on character-level RNN | DP-SGD · Surnames | accuracy: 60.86 -> 66.0 | Improved |
| Add individualized privacy assignment to individual accounting | Individual accounting · MNIST eps=0.3 | test accuracy: 93.64 -> 94.16 | Improved |
| Individual accounting alone vs vanilla DP-SGD | DP-SGD · MNIST eps=0.3 | test accuracy: 93.29 -> 93.64 | Improved |
Closest papers by meaning
- Differential Privacy Has Disparate Impact on Model Accuracy (NeurIPS 2019 · similarity 0.91)
- Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement (NeurIPS 2024 · similarity 0.89)
- Proactive DP: A Multiple Target Optimization Framework for DP-SGD (ICML 2024 · similarity 0.88)
- Nearly Tight Black-Box Auditing of Differentially Private Machine Learning (NeurIPS 2024 · similarity 0.88)
- It's My Data Too: Private ML for Datasets with Multi-User Training Examples (ICML 2025 · similarity 0.88)
- One-shot Empirical Privacy Estimation for Federated Learning (ICLR 2024 · similarity 0.88)
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.
