Research map / Federated learning and privacy / Privacy attacks and secure learning
SecMoE: Communication-Efficient Secure MoE Inference via Select-Then-Compute
Bowen Shen, Yuyue Chen, Peng Yang, Bin Zhang, Xi Zhang, Zoe L. Jiang
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 |
|---|---|---|---|
| select expert obliviously then compute only one expert (Select-Then-Compute) instead of evaluating all experts | secure MoE inference vs BumbleBee · MoE-Small LAN, 128 experts | runtime (min): 9.76 -> 2.52 | Improved |
| Select-Then-Compute MoE layer vs Iron | secure MoE inference · MoE-Small, 128 experts (communication) | communication (GB): 24.17 -> 0.82 | Improved |
| Select-Then-Compute vs BumbleBee on WAN | secure MoE inference · MoE-Small WAN, 128 experts | runtime (min): 13.88 -> 3.68 | Improved |
| increase experts from 32 to 128 in SecMoE (ablation) | SecMoE · MoE-Small LAN | runtime (min): 2.04 -> 2.52 | No clear effect |
| replace BumbleBee GeLU with Select-Then-Compute piecewise quadratic GeLU (ablation) | SecMoE · Switch-Base 128 experts | GeLU communication: 81% less communication, 7.1x speedup | Improved |
| replace BumbleBee GeLU with Select-Then-Compute GeLU (ablation) | SecMoE · MoE-Small 128 experts | GeLU communication / LAN runtime: 44% less communication, 11% less LAN runtime | Improved |
| use secure SecMoE protocol instead of plaintext inference | MoE-Small · CoLA/QNLI/RTE validation | MCC / accuracy: QNLI 90.3 -> 90.2; RTE 69.9 -> 70.0; CoLA 41.0 -> 41.0 | No clear effect |
Closest papers by meaning
- CryptoMoE: Privacy-Preserving and Scalable Mixture of Experts Inference via Balanced Expert Routing (NeurIPS 2025 · similarity 0.90)
- CipherPrune: Efficient and Scalable Private Transformer Inference (ICLR 2025 · similarity 0.89)
- Encryption-Friendly LLM Architecture (ICLR 2025 · similarity 0.88)
- United We Stand: Accelerating Privacy-Preserving Neural Inference by Conjunctive Optimization with Interleaved Nexus (AAAI 2024 · similarity 0.87)
- An Efficient Private GPT Never Autoregressively Decodes (ICML 2025 · similarity 0.87)
- Converting Transformers to Polynomial Form for Secure Inference Over Homomorphic Encryption (ICML 2024 · similarity 0.86)
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.
