Secure Association Rule Mining on Vertically Partitioned Data Using Private-Set Intersection

Abstract

Data mining entails the discovery of unexpected but reusable knowledge from large unorganized datasets. Among the many available data-mining algorithms, association rule mining (ARM) is very common. It was developed to aggregate all data into one site and subsequently mine them. In recent years, organizations in different fields have been required to collaborate to create new value. However, data mining among and within organizations has raised privacy and confidentiality concerns. In our scheme, parties cannot share anything other than the number of records, including the candidate itemset. This study focuses on the private-set intersection instead of the scalar product and shows that this intersection enables organizations to execute ARM on vertically partitioned data, allowing flexible information sharing while preserving privacy without increasing communication and computation costs. Furthermore, we focus on the fact that the number of protocol rounds among parties can be reduced and present three use cases in which the proposed scheme works more effectively than the existing schemes.