Research On Model And Algorithm For Privacy Preserving Clustering Based On Rule Hiding

In recent years, with the development of information technology, the operation of enterprises has gone through a drastic revolution. Data information flow became the lifeblood of enterprises. However, in such a situation, people would worry about disclosure of privacy and are likely to provide phony information rather than the authentic. So, the level of privacy preserving determines whether enterprises can get real information from which they could mine the useful rule. Upon that, privacy preserving became the focus of people's attention.In this dissertation, taxonomy of privacy preserving algorithm is developed. Based on this taxonomy, discover a vacancy of recent research which is a preserving clustering based on rule hiding. Subsequently, a novel model is built and an algorithm is put forward. In addition, introduce performance measures for privacy preserving and report the results. The tests show that such an approach is fairly effective. The contributions of this dissertation are as follows:First of all, taxonomy of privacy preserving algorithm is developed, so that the algorithms could be categorized and compared according to application types, technique strategy, hiding object and data mining algorithm. Therefore, a vacancy, preserving clustering based on rule hiding, is discovered.Secondly, a novel model is built and an algorithm is put forward about preserving clustering based on rule hiding, including: data preprocess, clustering rule hiding and algorithm evaluation. Above all, the raw data are processed to get the input variable of clustering rule hiding algorithm by normalization, k-means clustering algorithm and silhouette coefficient. Subsequently, the data are transformed by adding noise, so as to achieve the goal of privacy preserving. Three steps of the clustering rule hiding algorithm are as the following: (1) deciding sensitive objects; (2) choosing noises; (3) modifying data according to the noises. Finally, introduce the performance measures which are computational complexity, effectiveness of privacy preserving and accuracy.Thirdly, two sets of data are tested and the results are reported. The tests show that such an approach is fairly effective.