Privacy-Preserving Mining of Association Rules for Horizontally Distributed Databases Based on FP-Tree

Abstract

The discovery of frequent patterns, association rules, and correlation relationships among huge amounts of data is useful to business intelligence in this big data era. We propose a new scheme which is a secure and efficient association rule mining (ARM) method on horizontally partitioned databases. We enhance the performance of ARM on distributed databases by combining Apriori algorithm and FP-tree in this new situation. To help the implement of combining Apriori algorithm and FP-tree on distributed databases, we originally come up with a method of merging FP-tree in our scheme. We take advantage of Homomorphic Encryption to guarantee the security and efficiency of data operation in our scheme. More speficially, we use Paillier’s homomorphic encryption method which only has addition homogeneity to encrypt items’ supports. At last, we perform experimental analysis for our scheme to show that our proposal outperform the existing schemes.