Research map / Federated learning and privacy / Privacy attacks and secure learning
[Re] Privacy-preserving collaborative learning with automatic transformation search
Alfonso Taboada Warmerdam, Lodewijk Loerakker, Lucas Meijer, Ole Nissen
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 |
|---|---|---|---|
| apply the hybrid of searched transformation policies instead of no transformation | ResNet20 · CIFAR100, Adam+Cosine reconstruction attack | PSNR: 13.88 -> 7.64 | Improved |
| apply the hybrid of searched transformation policies instead of no transformation | ResNet20 · CIFAR100 | ACC: 76.88 -> 77.92 | Improved |
| apply the hybrid of searched transformation policies instead of no transformation | ConvNet · CIFAR100, Adam+Cosine reconstruction attack | PSNR: 13.07 -> 6.83 | Improved |
| apply the hybrid of searched transformation policies instead of no transformation | ConvNet · CIFAR100 | ACC: 70.13 -> 70.27 | Improved |
| apply the hybrid of searched transformation policies instead of no transformation | ResNet20 · F-MNIST, Adam+Cosine reconstruction attack | PSNR: 10.04 -> 7.60 | Improved |
| apply the hybrid of searched transformation policies instead of no transformation | ResNet20 · F-MNIST | ACC: 95.03 -> 92.23 | Hurt |
| apply the hybrid of searched transformation policies instead of no transformation | ConvNet · F-MNIST, Adam+Cosine reconstruction attack | PSNR: 9.12 -> 6.94 | Improved |
| apply the hybrid of searched transformation policies instead of no transformation | ConvNet · F-MNIST | ACC: 94.25 -> 91.35 | Hurt |
| use hybrid selection rather than one fixed searched policy (ablation) | ResNet20 · CIFAR100 | ACC: 70.56 -> 77.92 | Improved |
| use hybrid selection rather than one fixed searched policy (ablation) | ConvNet · CIFAR100 | ACC: 66.98 -> 70.27 | Improved |
| use hybrid selection rather than one fixed searched policy (ablation) | ResNet20 · F-MNIST | ACC: 89.41 -> 92.23 | Improved |
| use hybrid selection rather than one fixed searched policy (ablation) | ConvNet · F-MNIST | ACC: 90.56 -> 91.35 | Improved |
Closest papers by meaning
- Privacy-Preserving Collaborative Learning With Automatic Transformation Search (CVPR 2021 · similarity 0.91)
- Revisiting Gradient Pruning: A Dual Realization for Defending against Gradient Attacks (AAAI 2024 · similarity 0.82)
- Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearning (NeurIPS 2024 · similarity 0.80)
- Leveraging Per-Instance Privacy for Machine Unlearning (ICML 2025 · similarity 0.80)
- MaSS: Multi-attribute Selective Suppression for Utility-preserving Data Transformation from an Information-theoretic Perspective (ICML 2024 · similarity 0.80)
- Improved Bounds for Pure Private Agnostic Learning: Item-Level and User-Level Privacy (ICML 2024 · similarity 0.79)
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.
