Research map / Federated learning and privacy / Privacy attacks and secure learning
Partially Encrypted Deep Learning using Functional Encryption
Théo Ryffel, David Pointcheval, Francis Bach, Edouard Dufour-Sans, Romain Gay
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 |
|---|---|---|---|
| reduce private network output size from 10 to 4 (ablation) | partially encrypted polynomial net · synthetic two-font MNIST | collateral (font) accuracy: 93.5% -> 76.4% | Improved |
| reduce private network output size to 3 (ablation) | partially encrypted polynomial net · synthetic two-font MNIST | main digit accuracy: 97.7% -> 97.1% | Hurt |
| add semi-adversarial training against a simulated font adversary | private quadratic network · synthetic two-font MNIST | collateral (font) accuracy: 93.5% -> 53.5-58.9% across adversary models (near random) | Improved |
| use adversary models with different architecture than the one trained against (kNN, random forest) | adversarial training defense · font distinction | collateral accuracy: 58.1% (kNN), 58.9% (RF) vs 53.5% (ridge) | Hurt |
| quantize weights and input to 4 bits (ablation) | private network · synthetic two-font MNIST | main accuracy | No clear effect |
Closest papers by meaning
- Revisiting ML Training under Fully Homomorphic Encryption: Convergence Guarantees, Differential Privacy, and Efficient Algorithms (ICML 2026 · similarity 0.88)
- HEPrune: Fast Private Training of Deep Neural Networks With Encrypted Data Pruning (NeurIPS 2024 · similarity 0.85)
- ReBoot: Encrypted Training of Deep Neural Networks with CKKS Bootstrapping (AAAI 2026 · similarity 0.84)
- United We Stand: Accelerating Privacy-Preserving Neural Inference by Conjunctive Optimization with Interleaved Nexus (AAAI 2024 · similarity 0.84)
- FicGCN: Unveiling the Homomorphic Encryption Efficiency from Irregular Graph Convolutional Networks (ICML 2025 · similarity 0.83)
- DCT-CryptoNets: Scaling Private Inference in the Frequency Domain (ICLR 2025 · similarity 0.82)
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.
