NKAB: An optimization approach for k-anonymity based on Black Hole Algorithm

Abstract

This paper addresses the NP-hard problem of optimal k-anonymization. We propose NKAB, a novel optimization algorithm that significantly enhances the effectiveness of our earlier K-Anonymity Black Hole Algorithm (KAB). Unlike the original algorithm, NKAB introduces a preprocessing phase based on the concept of Natural Equivalence Classes, which filters and groups records with identical quasi-identifiers already present in the original dataset. This step significantly reduces search space and improves the computational efficiency of KAB. Experimental results obtained on the Adult dataset, a standard benchmark for the evaluation of anonymization algorithms, show a reduction in the search space, up to 53.5% for the privacy parameter $k=2$, leading to an average computation time reduction of 78.8%, while maintaining high data utility with lower information loss (ranging from 0.88% to 10.72%).