Efficient, Accurate and Privacy-Preserving Data Mining for Frequent Itemsets in Distributed Databases

Mining distributed databases is emerging as a fundamental computational problem. A common approach for mining distributed databases is to move all of the data from each database to a central site and a single model is built. This approach is accurate, but too expensive in terms of time required. For this reason, several approaches were developed to efficiently mine distributed databases, but they still ignore a key issue privacy. Privacy is the right of individuals or organizations to keep their own information secret. Privacy concerns can prevent data movement data may be distributed among several custodians, none of which is allowed to transfer its data to another site. In this paper we present an efficient approach for mining frequent itemsets in distributed databases. Our approach is accurate and uses a privacy-preserving communication mechanism. The proposed approach is also efficient in terms of message passing overhead, requiring only one round of communication during the mining operation. We show that our privacy-preserving distributed approach has superior performance when compared to the application of a well-known mining algorithm in distributed databases.