K−means Clustering Microaggregation for Statistical Disclosure Control

This paper presents a K-means clustering technique that satisfies the biobjective function to minimize the information loss and maintain k-anonymity. The proposed technique starts with one cluster and subsequently partitions the dataset into two or more clusters such that the total information loss across all clusters is the least, while satisfying the k-anonymity requirement. The structure of K−means clustering problem is defined and investigated and an algorithm of the proposed problem is developed. The performance of the K− means clustering algorithm is compared against the most recent microaggregation methods. Experimental results show that K−means clustering algorithm incurs less information loss than the latest microaggregation methods for all of the test situations.