Vertically Partitioned High-Dimensional Data Publishing with Differential Privacy

Le résumé fourni par la source

Ensuring data security and availability is crucial in high-dimensional data privacy releases, but the effectiveness of dimensionality reduction algorithms and the allocation of privacy budgets often pose challenges. Data usability encompasses accuracy, completeness, and association properties. To address these issues, the rough set-based personalized privacy-preserving Bayesian network algorithm (RSPPrivBayes) is proposed. This algorithm leverages the advantages of rough sets and Bayesian networks: first, it preprocesses the data using rough set theory to retain key correlations while eliminating redundancy for efficient dimensionality reduction; next, according to user privacy requirements, it differentially injects Laplace noise into the categorized data to balance privacy and data correlation; finally, it employs a Bayesian network to construct the released data, thereby protecting privacy while preserving dependency relationships and statistical properties. Experiments confirm that RSPPrivBayes performs well in both privacy protection and data quality, verifying its effectiveness and efficiency.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.