A Privacy Preserving Model in Uncertain Graph Mining

Abstract

Social network graph data provide rich research value and are collected and published by social network managers. Unfortunately, these data usually contain a large number of user sensitive information, which will cause privacy leakage to publish it. Recently, uncertain graph method has become an effective privacy preserving method for social network data publishing. Because it makes the original graph become a probability graph, the edge existence probability is between [0, 1] in it, which effectively prevents the background knowledge attack. Although some work has been done with uncertain graph method, they are generally limited in data utility, which is due to the fact that low probability edges have no research value. In this paper, we introduce a privacy preserving method for injecting uncertainty into community to generate an uncertain graph. Our key idea is to obfuscate the partial of original graph to achieve privacy preserving while effectively improving data utility. We validate the method and evaluate the utility of the generated uncertain graph using publicly real datasets.