Research map / Federated learning and privacy / Differential privacy
Differential Privacy Has Disparate Impact on Model Accuracy
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 |
|---|---|---|---|
| Apply DP-SGD instead of non-DP training | MNIST CNN · MNIST, class 8 | accuracy: 97% -> 77% | Hurt |
| Apply DP-SGD instead of non-DP training | MNIST CNN · MNIST, class 2 | accuracy: 99% -> 98% | Hurt |
| Add gradient clipping without noise (ablation) | MNIST DP-SGD · MNIST, class 8 | accuracy: 97% -> 93% | Hurt |
| Add gradient noise without clipping (ablation) | MNIST DP-SGD · MNIST, class 8 | accuracy: 97% -> 93% | Hurt |
| Add both gradient clipping and noise (ablation) | MNIST DP-SGD · MNIST, class 8 | accuracy: 93% -> 77% | Hurt |
| Add both gradient clipping and noise (ablation) | MNIST DP-SGD · MNIST, well-represented classes | accuracy: 99% -> 98% | Hurt |
| Use DP-SGD rather than non-DP training | Face gender classification · Darker-skinned faces | accuracy | Hurt |
| Use DP-SGD rather than non-DP training | Tweet sentiment classifier · AAE tweets | accuracy | Hurt |
| Use DP-SGD rather than non-DP training | iNaturalist species classifier · Underrepresented species classes | accuracy | Hurt |
| Use DP-FedAvg rather than non-DP federated training | Reddit language model · Participants with moderately sized vocabularies (500-1000 words) | accuracy | Hurt |
| Use DP-FedAvg rather than non-DP federated training | Reddit language model · Participants with extremely small vocabularies | accuracy | Helped |
| Increase batch size (ablation) | MNIST DP-SGD · MNIST class accuracy gap | accuracy gap | Helped |
Closest papers by meaning
- Optimal Differentially Private Model Training with Public Data (ICML 2024 · similarity 0.92)
- Towards hyperparameter-free optimization with differential privacy (ICLR 2025 · similarity 0.91)
- The Limits of Differential Privacy in Online Learning (NeurIPS 2024 · similarity 0.90)
- Nearly Tight Black-Box Auditing of Differentially Private Machine Learning (NeurIPS 2024 · similarity 0.90)
- Noise-Aware Differentially Private Regression via Meta-Learning (NeurIPS 2024 · similarity 0.90)
- Attack-Aware Noise Calibration for Differential Privacy (NeurIPS 2024 · similarity 0.90)
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.