Private Frequent Itemset Mining in the Local Setting.

Set-valued data, which is useful for representing user-generated data, becomes ubiquitous in numerous online services. Service provider profits by learning patterns and associations from users’ set-valued data. However, it comes with privacy concerns if these data are collected from users directly. This work studies frequent itemset mining from user-generated set-valued data meanwhile locally preserving personal data privacy. Under local d-privacy constraints, which capture intrinsic dissimilarity between set-valued data in the framework of differential privacy, we propose a novel privacy-preserving frequent itemset mining mechanism, called PrivFIM. It provides rigorous data privacy protection on the user-side and allows effective statistical analyses on the server-side. Specifically, each user perturbs his set-valued data locally to guarantee that the server cannot infer the user’s original itemset with high confidence. The server can reconstruct an unbiased estimation of itemset frequency from these randomized data and then combines it with the Apriori-based pruning technique to identify frequent itemsets efficiently and accurately. Extensive experiments conducted on real-world and synthetic datasets demonstrate that PrivFIM surpasses existing methods, and maintains high utility while providing strong privacy guarantees.