Secure two-party association rule mining

Association rule mining algorithm provides a means for determining rules and patterns from a large collection of data. However, when two sites want to engage in an association rule mining, data privacy concerns are raised. These concerns include loosing a competitive edge in the market place and breaching privacy laws. Techniques that have addressed this problem are data perturbation and homomorphic encryption. Homomorphic encryption based solutions produce more accurate results than data perturbation. Most previous solutions for privacy preserving association rule mining require the disclosure of intermediate mining results such as support counts and database size todetermine frequent itemset. To overcome this weakness we propose a secure comparison technique based on state-of-the-art fully homomorphic encryption scheme, by which we build secure twoparty association rule mining protocol. Our solution preserves complete privacy of both parties and it is more efficient than other solutions because there is no need for exponentiation of numbers.