Research map / Federated learning and privacy / Privacy attacks and secure learning
Alleviating Privacy Attacks via Causal Learning
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 |
|---|---|---|---|
| Use causal parents instead of all features in a DNN | Predictive model · Sachs (Akt) | attack accuracy (%): 76 -> 50 | Helped |
| Use IRM to learn invariant causal features instead of ERM | Colored MNIST model · Colored MNIST | attack accuracy (%): 66 -> 53 | Helped |
| Use IRM to learn invariant causal features instead of ERM | Colored MNIST model · Colored MNIST | test accuracy (%): 16 -> 69 | Helped |
| Use IRM instead of ERM | Colored MNIST model · Colored MNIST | train accuracy (%): 87 -> 70 | Hurt |
| Add 1 non-causal parent (ablation) | Causal model · Sachs (Akt) | attack accuracy (%): 50 -> 52 | Hurt |
| Add 2 non-causal parents (ablation) | Causal model · Sachs (Akt) | attack accuracy (%): 50 -> 61 | Hurt |
| Add 1 non-causal parent (ablation) | Causal model · Sachs (Akt) | prediction accuracy (%): 79 -> 75 | Hurt |
| Add 2 non-causal parents (ablation) | Causal model · Sachs (Akt) | prediction accuracy (%): 79 -> 68.8 | Hurt |
| Use causal models instead of DNNs under shifted test distributions | Bayesian network prediction models · Test(P*) across Bayesian network datasets | attack accuracy: 0.8; below 0.6 | Helped |
| Use causal models rather than associational models | Prediction models · Test(P*) across Bayesian network datasets | attack accuracy: 0.5 | Helped |
Closest papers by meaning
- Causal Evaluation of Membership Inference Attacks (NeurIPS 2026 · similarity 0.87)
- Assessing Per-Sample Membership Inference Vulnerability without Retraining (NeurIPS 2026 · similarity 0.86)
- Privacy Backdoors: Enhancing Membership Inference through Poisoning Pre-trained Models (NeurIPS 2024 · similarity 0.85)
- Instance-Specific Asymmetric Sensitivity in Differential Privacy (NeurIPS 2024 · similarity 0.85)
- Impact of Dataset Properties on Membership Inference Vulnerability of Deep Transfer Learning (NeurIPS 2025 · similarity 0.85)
- Improving the Variance of Differentially Private Randomized Experiments through Clustering (ICML 2025 · similarity 0.85)
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.