GN: A privacy preserving data publishing method based on generalization and noise techniques

Abstract

Generalization is a popular technique to realize k-anonymity. However, when the distribution of original data is uneven, generalization will distort the data greatly, which makes the anonymous data low utility. To address the problem, we propose a GN method, which limits the degree of generalization by adding noise tuples during anonymization. We also propose a GN-Bottom-up algorithm to achieve k-anonymity based on GN method. Experiments show that the GN method can generate anonymous data with less distortion and higher classification accuracy than generalization method.