Experimental analysis of a privacy-preserving scalar product protocol

The recent investigation of privacy-preserving data mining has been motivated by the growing concern about the privacy of individuals when their data is stored, aggregated, and mined for information. In an effort towards practical algorithms for privacy-preserving data mining solutions, we analyze and implement solutions to an important primitive: the privacy-preserving scalar product of two vectors held by different parties. Privacypreserving scalar products are an important component of privacy-preserving data mining algorithms, particularly when data is vertically partitioned between two or more parties. We examine a cryptographically secure privacypreserving data mining solution in different computational settings. Our experimental results show that in the absence of special-purpose hardware accelerators or practical optimizations, the computational complexity, rather than the communication complexity, is the performance bottleneck. We also evaluate several practical optimizations to improve the effi-