An Information-Driven Genetic Algorithm for Privacy-Preserving Data Publishing

Abstract

Due to the expanding requirements for data publishing and growing concerns regarding data privacy, the privacy-preserving data publishing (PPDP) problem has received considerable attention from research communities, industries, and governments. However, it is challenging to tackle the trade-off between privacy preservation and data quality maintenance in PPDP. In this paper, an information-driven genetic algorithm (ID-GA) is designed to achieve optimal anonymization based on attribute generalization and record suppression. In ID-GA, an information-driven crossover operator is designed to efficiently exchange information between different anonymization solutions; an information-driven mutation operator is proposed to promote information release during anonymization; a two-dimension selection operator is designed to identify the qualities of different anonymization solutions. Experimental results verify the advantages of ID-GA in terms of solution accuracy and convergence speed. Besides, the impacts of all the proposed components are verified.