Research map / Federated learning and privacy / Privacy attacks and secure learning
Circa: Stochastic ReLUs for Private Deep Learning
Zahra Ghodsi, Nandan Kumar Jha, Brandon Reagen, Siddharth Garg
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 |
|---|---|---|---|
| replace Delphi ReLU garbled circuit with Circa stochastic ReLU | ResNet/VGG · CIFAR-10/100, TinyImageNet | PI runtime: 6.32 -> 2.47 s (ResNet32) | Improved |
| add Circa on top of DeepReDuce-optimized networks | DeepReDuce ResNet18 · CIFAR-100 and TinyImageNet | PI runtime: 2.76 -> 1.65 s (1.7x) | Improved |
| stochastic ReLU truncation (PosZero/NegPass) applied to ResNet32 | ResNet32 · CIFAR-10 | accuracy: 92.43% -> 91.85% | Hurt |
| refactor ReLU as x*sign(x) with sign in garbled circuit (ablation) | Circa GC · GC size per ReLU | GC size reduction: 1.4x smaller | Improved |
| make sign stochastic (skip modulo overflow handling) (ablation) | Circa GC · GC size per ReLU | GC size reduction: 1.9x smaller | Improved |
| truncate stochastic sign inputs to 12 bits (ablation) | Circa GC · GC size per ReLU | GC size reduction: 4.7x savings over baseline | Improved |
| use PosZero instead of NegPass fault mode (ablation) | Circa on ResNet18 · CIFAR-100, TinyImageNet | allowed truncation bits: 1-2 extra bits | Improved |
Closest papers by meaning
- Convex Approximation of Two-Layer ReLU Networks for Hidden State Differential Privacy (NeurIPS 2025 · similarity 0.84)
- Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement (NeurIPS 2024 · similarity 0.84)
- United We Stand: Accelerating Privacy-Preserving Neural Inference by Conjunctive Optimization with Interleaved Nexus (AAAI 2024 · similarity 0.84)
- ReLU Strikes Back: Exploiting Activation Sparsity in Large Language Models (ICLR 2024 · similarity 0.83)
- Jaguar: Fast Private CNN Inference with Power-of-Two Homomorphic Arithmetic (NeurIPS 2026 · similarity 0.83)
- Rethinking DP-SGD in Discrete Domain: Exploring Logistic Distribution in the Realm of signSGD (ICML 2024 · similarity 0.83)
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.
