Privacy-preserving distributed mining of association rules on horizontally partitioned data

Abstract

Abstract-Data mining can extract important knowledge from large data collections-but sometimes these collections are split among various parties. Privacy concerns may prevent the parties from directly sharing the data and some types of information about the data. This paper addresses secure mining of association rules over horizontally partitioned data. The methods incorporate cryptographic techniques to minimize the information shared, while adding little overhead to the mining task.

Conditional Encrypted Mapping and Comparing Encrypted Numbers

“…Many interesting properties and computations, such as transaction classification or rule mining, involve evaluating a very large number of instances of GT [10,13]. Our improvements also apply to solving interval membership problems (reduced to GT in [2]).…”

Privacy Preserving Data Mining Research: Current Status and Key Issues

“…Kantarcioglu and Clifton [7] suggested that adopting a common framework for discussing privacy preservation will enable next generation data mining technology to make substantial advances in alleviating privacy concerns. Verykios et al [28] analyzed the state-ofthe-art, classified the proposed algorithms from five different dimensions: data distribution, data modification, data mining algorithm, data or rule hiding, and privacy preservation.…”

Efficient Classification from Multiple Heterogeneous Databases

“…Either way of distribution divides the rows or columns of a table into different parts. Distributed data mining approaches for horizontally partitioned data include meta-learning [3] that merges models built from different sites, and privacy preserving techniques including decision tree [8] and association rule mining [7]. Those for vertically partitioned data include association rule mining [12] and k-means clustering [13].…”

“…However, perfect integration of heterogeneous data sources is a very challenging problem, and it is often impossible to migrate one whole database to another site. In contrast, distributed data mining [3,7,8,12,13] aims at discovering knowledge from a dataset that is stored at different sites. But they focus on a homogeneous dataset (a single table or a set of transactions) that is distributed to multiple sites, thus are unable to handle heterogeneous relational databases.…”