Research map / Federated learning and privacy / Privacy attacks and secure learning
Active Membership Inference Test (aMINT): Enhancing Model Auditability with Multi-Task Learning.
Daniel DeAlcala, Aythami Morales, Julian Fierrez, Gonzalo Mancera, Ruben Tolosana, Javier Ortega-Garcia
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 |
|---|---|---|---|
| add middle-setup AAD instead of entry (ablation) | Active MINT ResNet50 · CIFAR-10 | MINT accuracy: 0.86 -> 0.89 | Improved |
| use output-setup AAD instead of entry (ablation) | Active MINT ResNet50 · CIFAR-10 | audited accuracy: 0.53 -> 0.39 | Hurt |
| use output-setup AAD instead of entry (ablation) | Active MINT ResNet50 · CIFAR-10 | MINT accuracy: 0.86 -> 0.85 | Hurt |
| use entry-setup AAD instead of middle (ablation) | Active MINT ResNet50 · GTSRB | MINT accuracy: 0.83 -> 0.86 | Improved |
| train MINT model jointly (Active) instead of post hoc (Passive) | ResNet50 · CIFAR-10 | MINT accuracy: 0.66 -> 0.86 | Improved |
| train MINT model jointly (Active) instead of post hoc (Passive) | ResNet50 · MNIST | MINT accuracy: 0.51 -> 0.83 | Improved |
| train MINT model jointly (Active) instead of post hoc (Passive) | ResNet50 · GTSRB | MINT accuracy: 0.61 -> 0.86 | Improved |
| train MINT model jointly (Active) instead of post hoc (Passive) | ResNet50 · Tiny ImageNet | MINT accuracy: 0.59 -> 0.80 | Improved |
| train MINT model jointly (Active) instead of post hoc (Passive) | ViT · CIFAR-10 | MINT accuracy: 0.60 -> 0.86 | Improved |
| train MINT model jointly with the audited model | ResNet50 · CIFAR-10 | audited accuracy: 0.55 -> 0.53 | Hurt |
| train MINT model jointly with the audited model | ResNet50 · Tiny ImageNet | audited accuracy: 0.17 -> 0.12 | Hurt |
| train MINT model jointly with the audited model | MobileNet · MNIST | audited accuracy: 0.97 -> 0.96 | Hurt |
Closest papers by meaning
- SMI: Statistical Membership Inference for Reliable Unlearned Model Auditing (NeurIPS 2026 · similarity 0.84)
- Causal Evaluation of Membership Inference Attacks (NeurIPS 2026 · similarity 0.82)
- Active Statistical Inference (ICML 2024 · similarity 0.82)
- What Really is a Member? Discrediting Membership Inference via Poisoning (NeurIPS 2025 · similarity 0.82)
- NEAR: A Training-Free Pre-Estimator of Machine Learning Model Performance (ICLR 2025 · similarity 0.81)
- $\texttt{dattri}$: A Library for Efficient Data Attribution (NeurIPS 2024 · similarity 0.81)
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.
