Achieving Privacy-preserving Computation on Data Grids

Abstract

This paper proposes a generic Grid privacy-preserving computation(G2PC) model which supports privacy-preserving data analysis and computation on multiple distributed datasets without compromising both the raw data privacy of Grid nodes and data statistics (intermediate result) privacy. The center of the design is our novel Data Privacy-Preserving Broker (D2PB) that combines the GSI (Grid Security infrastructure) with a number of cryptographic primitives. G2PC model requires neither one-to-all interactions among participating entities, nor reassignment of security parameters when membership or data changes. Therefore, it is efficient, scalable, and suited to large-scale Data Grid systems that are expected to host thousands of dynamic nodes. The privacy-preserving variance computation and privacy-preserving k-means clustering algorithm have been used as examples to demonstrate the efficacy and efficiency of our proposed framework.