Analysis of k-Anonymity for Homogeneity Attack
Debasis Mohapatra, Manas Ranjan Patra · 2014
The size of data is increasing exponentially day by day. It is as much important to provide privacy as to provide utility. Objective of privacy preserving data mining is to serve knowledge extraction with out leakage of private sensitive information, k-anonymity approach is a group based anonymization approach devised to achieve this objective. In this paper (α,k)-anonymization is discussed, that is a performance analysis metric, that finds the degree of association of sensitive value with an equivalence class. An algorithm is designed to return α value for each sensitive attribute. This paper represents a complete bipartite graph representation of (α,k)-anonymization, that can also generate accurate α values by implementing the algorithm on the graph. Performance analysis of k-anonymization is discussed with respect to the effects of homogeneity attack based on different parameters.