Research map / Adversarial robustness and security / Attacks on text and graphs
Scalable Membership Inference Attacks via Quantile Regression
Martin Andres Bertran, Shuai Tang, Aaron Roth, Michael Kearns, Jamie Heather Morgenstern, Steven Wu
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 |
|---|---|---|---|
| Use conditional quantile regression instead of a marginal threshold | Membership inference on CIFAR-10, CIFAR-100, and ImageNet-1k (ResNet-50 targets) · CIFAR-10 and CIFAR-100, 1% FPR | precision: 48.56% -> 62.95%; 58.81% -> 79.57% | Improved |
| Use conditional quantile regression instead of a marginal threshold | Membership inference on ImageNet-1k ResNet-50 · ImageNet-1k, 1% and 0.1% FPR | precision: 47.62% -> 97.45%; 46.81% -> 99.64% | Improved |
| Use quantile regression instead of LIRA shadow models | Membership inference on ImageNet-1k ResNet-50 · ImageNet-1k, 1% and 0.1% FPR | precision: 94.57% -> 97.45%; 98.73% -> 99.64% | Improved |
| Use quantile regression instead of LIRA shadow models | Membership inference on CIFAR-10 and CIFAR-100 ResNet-50 · CIFAR-10 and CIFAR-100, 1% and 0.1% FPR | precision: 83.00% -> 62.95%; 93.70% -> 64.48%; 96.20% -> 79.57%; 99.02% -> 85.41% | Hurt |
| Use one quantile regression model instead of LIRA ensembles | Tabular membership inference · OpenML and ACS NY | precision: 83.35% vs 82.73%; 88.05% vs 90.31%; 87.35% vs 89.62%; 86.54% vs 79.57%; 67.31% vs 69.23%; 56.35% vs 52.53%; 63.98% vs 67 | Mixed |
| Increase the number of LIRA shadow models | LIRA on OpenML and ACS NY · OpenML and ACS NY | precision | Mixed |
| Use a shadow model architecture that mismatches the target architecture | LIRA on CIFAR-10 and CIFAR-100 · CIFAR-10 and CIFAR-100 | precision | Hurt |
| Minimize quantile-regression pinball loss on public data | Membership inference attacks across the evaluated datasets · Image and tabular experiments | precision | Improved |
Closest papers by meaning
- Membership Inference Attacks on Diffusion Models via Quantile Regression (ICML 2024 · similarity 0.90)
- Low-Cost High-Power Membership Inference Attacks (ICML 2024 · similarity 0.87)
- Chameleon: Increasing Label-Only Membership Leakage with Adaptive Poisoning (ICLR 2024 · similarity 0.87)
- Causal Evaluation of Membership Inference Attacks (NeurIPS 2026 · similarity 0.86)
- Confidence Is All You Need for MI Attacks (Student Abstract) (AAAI 2024 · similarity 0.86)
- Membership Inference Attacks against Fine-tuned Large Language Models via Self-prompt Calibration (NeurIPS 2024 · similarity 0.86)
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.
