CFM: collusion-free model of privacy preserving frequent itemset mining

Abstract

Although many privacy preserving frequent itemset mining protocols have been proposed to preserve the privacy of participants, most of them are vulnerable against collusion. Usually, these protocols are designed for semi-honest model; where in this model, it is assumed that the participants do not deviate from the protocol. However, in real world, participants may collude with each other in order to falsify the protocol or to obtain the secret values of other parties. In this paper, we analyse the vulnerability of previous privacy preserving frequent itemset mining protocols from privacy point of view, and then, we proposes a new protocol (CFM), which preserves the privacy of participants, even in collusion state. CFM is designed for mining frequent itemsets from homogenous (horizontally partitioned) data, which not only preserves the privacy of participants in collusion states, but also shows better performance in comparison with previous works. In order to achieve this goal, CFM employs a new secret sharing and secret summation scheme, which distributes secret values among participants. Privacy preserving level of CFM is evaluated based on the disclosure of sensitive information.