An end-to-end knowledge graph solution to the frequent itemset hiding problem

Abstract

The Frequent Itemset Hiding Problem (FIHP) constitutes a critical aspect of Privacy-Preserving Data Mining, aiming to protect sensitive information while extracting valuable patterns. Despite the ubiquity of Knowledge Graphs (KGs) in many domains, their integration with FIHP remains an underexplored research area. In this paper, we present a new solution to FIHP, leveraging graph-based algorithms and Community Detection. Our method employs the efficiency and structure of KGs to quickly locate what needs to be hidden in order to strategically conceal sensitive knowledge within the data, while ensuring the preservation of KG structure and utility. Through comprehensive evaluations on established FIHP datasets, we showcase the efficacy of our approach in terms of processing time and increased data privacy. The results imply that our proposed method holds promise for facilitating privacy-preserving KG analytics in real-world applications, particularly in scenarios involving substantial streaming data, where existing approaches encounter limitations.