Implementation, Optimization and Performance Tests of Privacy Preserving Mechanisms in Homogeneous Collaborative Association Rules Mining

摘要

This article focuses on optimization and performance tests of association rules multiparty mining algorithms. We present how to improve Secure Set Union performance by using a Common Decrypting Key for a commutative encryption. We have also improved Secure Set Union making it fully secure in a semi-honest model. As an example of the above mentioned mechanisms application, the article presents new algorithms of mining association rules on horizontally partitioned data with preserving data privacy - CDKSU (Secure Union with Common Decrypting Key) and its fully secure version - CDKRSU (Secure Union with Common Decrypting Key and secure duplicate Removing). Those algorithms are compared with KCS scheme since they are all based on FDM. As far as the performance optimization is concerned, the application of Elliptic Curve Cryptography vs Exponential Cryptography is presented as well. The real, working system implementing given algorithms is subjected to performance tests which results are presented and analyzed.