Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TAMI-MPC:Trusted Acceleration of Minimal-Interaction MPC for Efficient Nonlinear Inference

Zhuoran Li, Hanieh Totonchi Asl, Yifei Cai, Ebrahim Nouri, Danella Zhao

Abstract

Secure multi-party computation (MPC) offers a practical foundation for privacy-preserving machine learning at the edge. However, current MPC systems rely heavily on communication and computation-intensive primitives-such as secure comparison for nonlinear inference, which are often impractical on resource-constrained platforms. To enable real-time inference under a resource-constrained platform, we introduce a Trusted Acceleration of Minimal-Interaction MPC framework, TAMI-MPC, for nonlinear evaluation. Specifically, we reduce communication cost by redesigning the core primitives, leaf comparison, and tree merge, reducing the interactive round from log(n) to just 1 per operation. Furthermore, unlike prior work that heavily relies on oblivious transfer (OT), a well-known computational bottleneck, we leverage synchronized seeds inside the TEE to eliminate OT for the vast majority of our designs, along with a correlated-randomness reuse technique that keeps new designs computationally lightweight. To fully realize the potential, we design a specialized accelerator that restructures the dataflow across stages to enable continuous, fine-grained streaming and high parallelism, reducing memory overhead. Our design achieves up to 4.86x speedup on ResNet-50 inference, compared with state-of-the-art CNN frameworks, and achieves up to 7.44x speedup on BERT-base inference, compared with state-of-the-art LLM frameworks.

TAMI-MPC:Trusted Acceleration of Minimal-Interaction MPC for Efficient Nonlinear Inference

Abstract

Secure multi-party computation (MPC) offers a practical foundation for privacy-preserving machine learning at the edge. However, current MPC systems rely heavily on communication and computation-intensive primitives-such as secure comparison for nonlinear inference, which are often impractical on resource-constrained platforms. To enable real-time inference under a resource-constrained platform, we introduce a Trusted Acceleration of Minimal-Interaction MPC framework, TAMI-MPC, for nonlinear evaluation. Specifically, we reduce communication cost by redesigning the core primitives, leaf comparison, and tree merge, reducing the interactive round from log(n) to just 1 per operation. Furthermore, unlike prior work that heavily relies on oblivious transfer (OT), a well-known computational bottleneck, we leverage synchronized seeds inside the TEE to eliminate OT for the vast majority of our designs, along with a correlated-randomness reuse technique that keeps new designs computationally lightweight. To fully realize the potential, we design a specialized accelerator that restructures the dataflow across stages to enable continuous, fine-grained streaming and high parallelism, reducing memory overhead. Our design achieves up to 4.86x speedup on ResNet-50 inference, compared with state-of-the-art CNN frameworks, and achieves up to 7.44x speedup on BERT-base inference, compared with state-of-the-art LLM frameworks.

Paper Structure

This paper contains 19 sections, 7 equations, 10 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Background and Security Definition
  3. Secure Comparison (Millionaires' Protocol)
  4. Security Definition
  5. TAMI-MPC Secure Comparison $\mathcal{F}_{\text{Mill}}$
  6. TEE-assisted Leaf Comparison $\mathcal{F}_{\text{Comp}}$
  7. TEE-assisted Tree Merge $\mathcal{F}_{\text{PolyMult}}$
  8. Further Optimization
  9. TAMI-MPC Hardware Architecture
  10. Overall Architecture
  11. Pipeline-Aware Interleaved Leaf Comparison $\mathcal{F}_{\text{Comp}}$
  12. Memory-Efficient Management for Tree Merge $\mathcal{F}_{\text{PolyMult}}$
  13. Experiments
  14. Experiment Setup
  15. Analysis of $\mathcal{F}_{\text{Comp}}$ and $\mathcal{F}_{\text{PolyMult}}$
  16. ...and 4 more sections

Figures (10)

  • Figure 1: ResNet-50 inference performance with Cheetah huang2022cheetah.
  • Figure 2: Illustration of Millionaires' protocol.
  • Figure 3: Comparison of security boundary between state-of-the-art TEE-based works zhou2022ppmlaczhou2022efficient and TAMI-MPC.
  • Figure 4: Design of $\mathcal{F}_{\text{Comp}}$
  • Figure 5: Design of $\mathcal{F}_{\text{PolyMult}}$
  • ...and 5 more figures