Abstract: Aiming at the low implementation efficiency and poor availability of data published in dynamic social network as well as processing large-scale dynamic social network graphs in a stand-alone workstation environment, this paper proposes a Distributed Dynamic Social Network Privacy-Preserving method Based on Link Prediction（D-DSNBLP）. This method realizes the parallel processing of anonymous large-scale graph data by the iterative updating model of message in Pregel-like. Firstly, the node is grouped by fast iteration. Secondly, the candidate node set is constructed in parallel according to the node attribute value in each group. Finally, a mutex set is built to add edges, to protect the privacy of nodes. Experiments show that the D-DSNBLP method improves the efficiency of processing large-scale dynamic social network and ensures the data availability of anonymous graphs.

Key words: distributed, dynamic social network, privacy protection, Pregel-like, link prediction

摘要： 针对单机工作站环境下处理大规模动态社会网络图时执行效率低，以及动态社会网络发布中数据可用性较差的问题，提出基于预测链接的分布式动态社会网络隐私保护方法D-DSNBLP。该方法通过Pregel-like消息迭代更新模型，实现匿名大规模图数据的并行处理。首先通过快速迭代完成结点分组；其次根据各个组内的结点属性值并行构建候选结点集合；最后通过构建互斥边集合添加边，实现结点的隐私保护。实验表明，D-DSNBLP方法提高了大规模动态社会网络发布的效率，保证了匿名图的数据可用性。

关键词: 分布式, 动态社会网络, 隐私保护, Pregel-like, 预测链接