A new method for privacy preserving association rule mining using homomorphic encryption with a secure communication protocol

With the enormous amount of data growing exponentially, data owners aim to share data with each other to acquire an enhanced analytic view of their data. Distributed association rule mining over cloud computation helps data owners to extract knowledge from multiple databases. However, privacy is an important concept for data owners who want to share data and extract knowledge from aggregated data. This study proposes an outsourcing method to securely share and mine association rules from multiple parties protecting all parties' privacy. In this research, we utilize the properties of homomorphic encryption and propose a custom secure communication protocol. In our study, privacy of the shared data of all parties is ensured and we will not perform decryption in any step of the proposed method. We show that previous schemes could not properly guarantee the integrity of messages, furthermore, we propose our innovative communication protocol that uses an integrity checking function based on homomorphic encryption to authorize each party's shared data. We have implemented our scheme and showed even though an additional communication burden exists in our method, this method outperforms its competitors in terms of elapsed time and security. We have achieved more security especially in confidentiality and integrity of messages due to use of our proposed secure communication protocol. We also measured our proposed method's precision of mining parameters with different datasets to show the accuracy of our method on different scales will not change.