Research map / Federated learning and privacy / Privacy attacks and secure learning
Posthoc privacy guarantees for collaborative inference with modified Propose-Test-Release
Abhishek Singh, Praneeth Vepakomma, Vivek Sharma, Ramesh Raskar
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 contrastive and noisy regularization to ARL (ablation) | ARL obfuscator · MNIST eps=1 | test accuracy: 0.329 -> 0.514 | Improved |
| add contrastive and noisy regularization to ARL (ablation) | ARL obfuscator · FMNIST eps=1 | test accuracy: 0.214 -> 0.371 | Improved |
| add contrastive and noisy regularization to ARL (ablation) | ARL obfuscator · UTKFace eps=1 | test accuracy: 0.605 -> 0.650 | Improved |
| add noisy regularization to ARL only (ARL-N, no contrastive) (ablation) | ARL obfuscator · MNIST eps=10 | test accuracy: 0.882 -> 0.325 | Hurt |
| add adversarial representation learning to C-N (ablation) | C-N obfuscator · FMNIST eps=10 | test accuracy: 0.599 -> 0.695 | Improved |
| add adversarial representation learning to C-N (ablation) | C-N obfuscator · MNIST eps=10 | test accuracy: 0.913 -> 0.912 | No clear effect |
| add adversarial representation learning to C-N (ablation) | C-N obfuscator · UTKFace eps=10 | test accuracy: 0.722 -> 0.700 | Hurt |
| replace ARL-C-N with plain Encoder at high privacy budget | Obfuscator · MNIST eps=10 | test accuracy: 0.921 -> 0.912 | No clear effect |
| replace Encoder with ARL-C-N at strict privacy | Obfuscator · MNIST eps=1 | test accuracy: 0.428 -> 0.514 | Improved |
| add noisy regularization alone to plain pipeline vs ARL (ablation) | Obfuscator · FMNIST eps=5 | test accuracy: 0.557 -> 0.310 | Hurt |
| increase adversarial weighting alpha from 0 to 0.9999 (ablation) | Obfuscator · Local Lipschitz constant | local Lipschitz constant | Improved |
| use larger privacy budget eps (1 -> 10) | ARL-C-N · MNIST | test accuracy: 0.514 -> 0.912 | Improved |
Closest papers by meaning
- United We Stand: Accelerating Privacy-Preserving Neural Inference by Conjunctive Optimization with Interleaved Nexus (AAAI 2024 · similarity 0.87)
- Theoretically Unmasking Inference Attacks Against LDP-Protected Clients in Federated Vision Models (ICML 2025 · similarity 0.85)
- Reimagining Mutual Information for Enhanced Defense against Data Leakage in Collaborative Inference (NeurIPS 2024 · similarity 0.85)
- Certification for Differentially Private Prediction in Gradient-Based Training (ICML 2025 · similarity 0.85)
- All Rivers Run to the Sea: Private Learning with Asymmetric Flows (CVPR 2024 · similarity 0.85)
- Efficient and Private Marginal Reconstruction with Local Non-Negativity (NeurIPS 2024 · similarity 0.84)
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.
