The Preservation of Weighted Graphs based on Shuffle differential privacy in Social Networks

With the prevalence of big data and network technology, online social networks have exploded in popularity and have been constructed to the graph structure, especially weighted graphs. Weighted values contain some sensitive information of individuals. Naturally, publishing weighted values on graphs brings a great challenge concerning the privacy protection of individual. In this paper, this challenge is addressed by proposing an algorithm based on shuffle differential privacy-WGSM. The original graph is decomposed into several sub-graphs and the weighted sequence is generated for each sub-graph. Then, the weighted sequence is divided into a set of partitions, where the perturbations will be made. Next, Laplace noise is added into the weighted values of each partition and the perturbed weighted values are shuffled randomly. Finally, all the perturbed sub-graphs merge into a synthetic weighted graph to publish. Theoretical and experimental analysis prove that WGSM algorithm satisfies e-differential privacy and has better effectiveness compared with the DWTDP algorithm and the WGLDP algorithm.