An improved l-diversity model for numerical sensitive attributes

Abstract

L-diversity model is an effective model to thwart homogeneity attack and background knowledge attack for microdata, but it has some defects on handling numerical sensitive attributes. The paper proposes an improved l-diversity model to address the problem. The model divides numerical sensitive values into several levels, and realizes sensitive attribute l-diversity based on these levels. Distinct diversity and entropy diversity are defined. Based on these definitions, an l-incognito algorithm is designed to implement the improved model. Experimental results show that the improved l-diversity model can protect numerical sensitive attributes effectively, and the anonymity tables generated by the l-incognito algorithm have high sensitive attributes diversity, so can resist homogeneity attack and partial background knowledge attack effectively.