An Improved Privacy Preserving Algorithm for Publishing Social Network Data

With the rapid growth of social networks, privacy issues have been raised for using or sharing data including user's information. Simply removing the identities of the vertices before publishing the social network data is considered an ill-advised practice due to privacy concerns, because the structure of the graph itself can reveal the identities of individuals. To mitigate this problem, a so-called graph k-degree anonymous method has been proposed in recent studies, in which the structure of the original graph is modified to ensure that there are at least k nodes having the same degree for each degree number in the modified graph. However, this approach arouses two key issues: the protecting algorithm has to be efficient especially for conducting on a great amount of graph data, and the modifications of the original graph should be minimized for keeping the data utility. To deal with these issues, we introduce a measurement for the modification cost of anonymizing a graph, and devise a novel privacy preserving algorithm for publishing social network data based on a new degree sequence partition algorithm. We conduct the algorithm on several real-world datasets. The experimental results show that the algorithm improve the performance of anonymizing the social network data and reduced the modifications of the original graph data.