Research map / Federated learning and privacy / Differential privacy
Beyond Membership: Limitations of Add/Remove Adjacency in Differential Privacy
Gauri Pradhan, Joonas Jälkö, Santiago Zanella-Beguelin, Antti Honkela
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 |
|---|---|---|---|
| audit with substitute-adjacency worst-case dataset canaries instead of relying on add/remove accounting | DP-SGD · worst-case crafted dataset | empirical epsilon | Improved |
| use crafted gradient canaries | DP-SGD fine-tuning ViT-B-16 · CIFAR10 | empirical epsilon | Improved |
| use input-space canaries | DP-SGD fine-tuning ViT-B-16 · CIFAR10 | empirical epsilon | Mixed |
| lower the subsampling rate q (ablation) | input-space canaries · CIFAR10 fine-tuning | audit efficacy | Hurt |
| use input-space canaries on a model trained from scratch | DP-Adam MLP · Purchase100 | empirical epsilon | Hurt |
| use crafted gradient canaries on a model trained from scratch | DP-Adam MLP · Purchase100 | empirical epsilon | Improved |
| use gradient canaries | DP-SGD Sentence-BERT linear layer · SST-2 | empirical epsilon | Improved |
| add the canary to the training data | DP-SGD models · CIFAR10, Purchase100, SST-2 | test accuracy | No clear effect |
Closest papers by meaning
- A Unified Framework for Adversary-Aware Differential Privacy Bounds (NeurIPS 2026 · similarity 0.88)
- Auditing Private Prediction (ICML 2024 · similarity 0.87)
- Beyond the Calibration Point: Mechanism Comparison in Differential Privacy (ICML 2024 · similarity 0.87)
- Instance-Specific Asymmetric Sensitivity in Differential Privacy (NeurIPS 2024 · similarity 0.87)
- Unifying Re-Identification, Attribute Inference, and Data Reconstruction Risks in Differential Privacy (NeurIPS 2025 · similarity 0.87)
- Provable Privacy with Non-Private Pre-Processing (ICML 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.
