Research On The Key Techniques For Privacy Preserving On Weighted Social Networks

With the development of Internet technology, social networks generate massive amounts of data all the time. Massive amounts of data are convenient for researchers with data analysis and knowledge mining, but at the same time it will expose users’ privacy. Social network data privacy protection has been the hot issue among scholars in the field of data privacy research. An attacker can use a variety of background knowledge to attack against privacy. Most of the present technology on anonymity social network graphs can only deal with simple graphs, but cannot be applied to weighted graphs. In order to solve the above problems, privacy preserving technologies on weighted social networks are proposed.An analysis on path privacy disclosure problem in the weighted graph is given. Different privacy preserving approaches are proposed based on simple paths set and complex paths set for shortest path identification. Global_Generaliation algorithm for simple paths set is proposed. Generalization for simple paths set between the target nodes in weighted graph, unifies the weight range for each path in the paths set to satisfy anonymity, and also makes the generalization weight range including the true value to improve the availability of anonymous data. For complex paths set, Local_Optimization algorithm is proposed. Based on the global generalization, the overlapping edges of the path are optimized locally. Furthermore, the weighted graph k-possible path anonymity (KPPA for short) model is proposed to protect against shortest-path-based attacks.For the sensitive label privacy disclosure problem in weighted graph, the k-histogram-inverse-l-diversity (KH-inv-LD for short) model is proposed to protect weight sequence of nodes. Based on the label generalization, the inverse l-diversity is proposed to protect the sensitive label information. For the single sensitive attributes in k-histogram weight anonymous group, the SSAG algorithm is proposed. Moreover, this paper firstly considers the multi-sensitive-attribute privacy preserve in weighted social networks and presents an efficient MSAG algorithm to protect the sensitive labels.Extensive experiments on real data sets show that the algorithm performs well to efficiently protect the privacy in weighted graph. Meanwhile, it can retain the structural properties of the original graph and increase the utility of the weight information.