Paper 2022/1271

Privacy-preserving Federated Singular Value Decomposition

Abstract

Modern Singular Value Decomposition (SVD) computation dates back to the 1960s when the basis for the eigensystem package and linear algebra package routines was created. Since then, SVD has gained attraction and been widely applied in various scenarios, such as recommendation systems and principal component analyses. Federated SVD has recently emerged, where different parties could collaboratively compute SVD without exchanging raw data. Besides its inherited privacy protection, noise injection could be utilized to further increase the privacy guarantee of this privacy-friendly technique. This paper advances the state-of-science by improving an existing Federated SVD scheme with two-fold contributions. First, we revise its privacy guarantee in terms of Differential Privacy, the de-facto data privacy standard of the 21st century. Second, we increase its utility by reducing the added noise, which is achieved by employing Secure Aggregation, a cryptographic technique to prevent information leakage. Using a recommendation system use-case with real-world data, we demonstrate that our scheme outperforms the state-of-the-art Federated SVD solution.