Secure frequent itemset mining from horizontally distributed databases

摘要

We propose a protocol for hiding the infrequent itemsets in horizontally distributed databases. Databases are horizontally partitioned when they are distributed among different players. Each player holds identical schema, but have information on different objects. Mining of frequent itemsets existed in databases of different players cause sensitive data leakage from one player to another player or to any third party. Mining should not reveal any players locally supported frequent itemsets among themselves or any third party. In order to maintain privacy, security is needed. We can provide the partial security to the item sets by removing infrequent subsets from the candidate item sets. In this thesis, a protocol for deriving frequent itemsets from horizontally distributed databases which does not leak secret information of the participating players in mining has been implemented. This protocol uses Fast Distributed Mining (FDM) algorithm which is given by Cheungetalet.el[3]. FDM algorithm uses the technique of Apriori[1] Algorithm to mine the frequent itemsets from distributed environment.