Explore a conference talk from USENIX ATC '23 that introduces SecretFlow-SPU, a cutting-edge framework for privacy-preserving machine learning (PPML). Discover how this innovative solution addresses the challenges of developing efficient PPML programs using secure multi-party computation (MPC) techniques. Learn about the framework's unique approach, which allows users to run machine learning programs from various frameworks with minimal modifications while maintaining privacy. Gain insights into SecretFlow-SPU's architecture, including its frontend compiler and backend runtime, and understand how it optimizes code for MPC protocols. Examine the performance comparisons with other state-of-the-art MPC-enabled PPML frameworks, showcasing SecretFlow-SPU's superior speed and efficiency across different experimental settings, particularly in wide area networks.
SecretFlow-SPU - A Performant and User-Friendly Framework for Privacy-Preserving Machine Learning
Overview
Syllabus
USENIX ATC '23 - SecretFlow-SPU: A Performant and User-Friendly Framework for Privacy-Preserving...
Taught by
USENIX
Related Courses
-
MD-ML: Super Fast Privacy-Preserving Machine Learning for Malicious Security with Dishonest Majority
-
EnvPipe - Performance-Preserving DNN Training Framework for Saving Energy
-
SecureML - A System for Scalable Privacy-Preserving Machine Learning
-
Prime Match - A Privacy-Preserving Inventory Matching System
-
Privacy-Preserving Machine Learning
-
Privacy-Preserving Machine Learning Workshop - Invited and Contributed Talks
