The Communication-Friendly Privacy-Preserving Machine Learning against Malicious Adversaries

TL;DR

This paper introduces an efficient GPU-accelerated maliciously secure MPC protocol for privacy-preserving machine learning, capable of handling complex models like CNNs with improved performance and scalability.

Contribution

It presents a novel, efficient protocol for secure linear function evaluation in malicious MPC, extended to non-linear layers, and implemented on GPUs for scalable privacy-preserving ML.

Findings

Significant efficiency improvements in malicious MPC on GPUs.

Compatibility with complex ML models like CNNs.

Effective secure inference for various machine learning workflows.

Abstract

With the increasing emphasis on privacy regulations, such as GDPR, protecting individual privacy and ensuring compliance have become critical concerns for both individuals and organizations. Privacy-preserving machine learning (PPML) is an innovative approach that allows for secure data analysis while safeguarding sensitive information. It enables organizations to extract valuable insights from data without compromising privacy. Secure multi-party computation (MPC) is a key tool in PPML, as it allows multiple parties to jointly compute functions without revealing their private inputs, making it essential in multi-server environments. We address the performance overhead of existing maliciously secure protocols, particularly in finite rings like $\mathbb{Z}_{2^\ell}$, by introducing an efficient protocol for secure linear function evaluation. We implement our maliciously secure MPC protocol on GPUs, significantly improving its efficiency and scalability. We extend the protocol to handle linear and non-linear layers, ensuring compatibility with a wide range of machine-learning models. Finally, we comprehensively evaluate machine learning models by integrating our protocol into the workflow, enabling secure and efficient inference across simple and complex models, such as convolutional neural networks (CNNs).