Discovering multivariate linear relationship securely

Abstract

This paper considers the privacy-preserving cooperative linear system of equations (PPC-LSE) problem in a large, heterogeneous, distributed database scenario. It proposes a privacy-preserving algorithm to discover multivariate linear relationship based on factor analysis. Compared with other PPC-LSE algorithms, the proposed algorithm not only significantly reduces the communication cost, but also avoids the random matrix generation of either party to hide private information.