An efficient cacheable secure scalar product protocol for privacy-preserving data mining

Abstract

Computing scalar products amongst private vectors in a secure manner is a frequent operation in privacy-preserving data mining algorithms, especially when data is vertically partitioned on many parties. Existing secure scalar product protocols based on cryptography are costly, particularly when they are performed repeatedly in privacy-preserving data mining algorithms. To address this issue, we propose an efficient cacheable secure scalar product protocol called CSSP that is built upon a homomorphic multiplicative cryptosystem. CSSP allows one to reuse the already cached data and thus, it greatly reduces the running time of any privacy-preserving data mining algorithms that adopt it. We also conduct experiments on real-life datasets to show the efficiency of the protocol. © 2011 Springer-Verlag.