An Privacy Protection Algorithm Based On K-Anonymity

Data publishing provides convenience for data exchange and date sharing, but personal privacy disclosure issues become also increasingly prominent in the process of data publishing. Privacy protection has become a new challenge of the database security field. As an important method of protecting personal privacy for data publishing, K-anonymity receives a wide range of concerns. This paper deeply studies K-anonymity technology, and in order to better balance contradictions problems between the privacy preservation and availability of anonymous data, a kind of new anonymity model and algorithm are proposed. The main contents are as follows.According to the current K-anonymity privacy protection models don't fully take into account privacy protection degree of sensitive attribute, the paper proposes a novel model (p,α)-Sensitive K-anonymity based on grouping of privacy protection degree of sensitive attribute. The model analyzes different values of sensitive attribute firstly, and divides sensitive attribute into groups according to the privacy protection degree, then sets different group privacy disclosure rate for each group. Thus, the idea not only provides the same protection for the same privacy protection degree of sensitive attribute, but also supplies better protection for attribute value of the higher sensitivity. At the same time, the paper analyses the current shortage of anonymous generalization algorithm, and implements (p,α)-Sensitive K-anonymity model combining the clustering with generalization. In the process of generating clustering, each tuple of clustering is as similar to each other as possible, and the corresponding distance definition and information loss calculation formula as well as the clustering generalization algorithm are given.The paper tests the model with classical Adult data set, and analyses the algorithm from two aspects of executive time and information loss. The experimental results suggest that the proposed solution can not only effectively protect privacy information of the higher sensitivity, reduce privacy disclosure risk apparently, but also reduce anonymous information loss, improve the quality of data.