On Association Rules Mining Algorithms with Data Privacy Preserving

Abstract

Data privacy becomes more and more important in data mining models designing. The concept of privacy preserving when performing data mining in distributed environment assumes that none of the databases shares its private data with the others. In our paper we analyze efficiency of two algorithms of privacy association rule mining in distributed data base. The algorithms are: HPSU (Horizontal Partitioning Secure Union) using horizontally partitioned database and VPSI (Vertical Partitioning Secure Intersection) using vertically partitioned database. To protect private data, HPSU uses secure union, and VPSI uses secure intersection. We implemented a system automatically per-forming analyses of these two algorithms using the same data. We point out possibilities of modifying the algorithms and discus the impact of these modifi-cations on the data privacy level.