A 30.4GOPS/mW MK-CKKS Processor for Secure Multi-Party Computation

📋 论文概要

本文提出了一款基于MK-CKKS方案的安全多党计算处理器，实现了30.4GOPS/mW的高能效，解决了多用户协作计算中的数据隐私保护问题。

💡 主要创新点

• 首次将MK-CKKS方案硬件化，支持无限数量参与方的安全计算
• 通过架构优化实现30.4GOPS/mW的能效，显著优于现有方案

🏷 关键词

📄 原文摘要

Secure data processing has become critical to privacy-preserving in the data-driven AI era. Multi-party computation (MPC) enables computations among multiple parties (users) in a collaborative way while preserving data privacy [1, 2], as illustrated in Fig. 17.3.1. Homomorphic encryption (HE), which enables computations on encrypted data, is a fundamental component of MPC. There are three major schemes to implement MPC: HE with secret sharing [3], fully HE (FHE) [4], and multi-key FHE (MK-FHE) [5]. Among these schemes, MK-FHE is the first scheme that can support an unlimited number of parties. It also allows parties to join or leave the computation on the fly. Moreover, MK-FHE enables non-interactive communication between parties to reduce the need for frequent data exchange. In MK-FHE, MK-CKKS [6] is a robust algorithm specifically for fixed-point data. In MK-CKKS, a designated party handles homomorphic operations on encrypted data. The

👥 作者与机构

Liang-Hsin Lin, Yao-Kai Yang, Chia-Hsiang Yang

National Taiwan University, Taipei, Taiwan