图数据发布隐私保护的聚类匿名方法

发布时间：2018-11-14 19:00

【摘要】：社交网络中积累的海量信息构成一类图大数据,为防范隐私泄露,一般在发布此类数据时需要做匿名化处理.针对现有匿名方案难以防范同时以结构和属性信息为背景知识的攻击的不足,研究一种基于节点连接结构和属性值的属性图聚类匿名化方法,利用属性图表示社交网络数据,综合根据节点间的结构和属性相似度,将图中所有节点聚类成一些包含节点个数不小于k的超点,特别针对各超点进行匿名化处理.该方法中,超点的子图隐匿和属性概化可以分别防范一切基于结构和属性背景知识的识别攻击.另外,聚类过程平衡了节点间的连接紧密性和属性值相近性,有利于减小结构和属性的总体信息损失值,较好地维持数据的可用性.实验结果表明了该方法在实现算法功能和减少信息损失方面的有效性.
[Abstract]:The huge amount of information accumulated in social networks constitutes a kind of map big data. In order to prevent privacy leaks, anonymous processing is usually needed when publishing this kind of data. Because the existing anonymous scheme is difficult to prevent the attack with structure and attribute information as the background knowledge, a method of clustering anonymous attribute graph based on node connection structure and attribute value is studied, and the attribute graph is used to represent the social network data. According to the structure and attribute similarity of nodes, all nodes in the graph are clustered into some superpoints containing the number of nodes not less than k, especially for the anonymous processing of each superpoint. In this method, the subgraph concealment and attribute generalizability of hyperpoints can prevent all recognition attacks based on structure and attribute background knowledge respectively. In addition, the clustering process balances the connection tightness between nodes and the similarity of attribute values, which is helpful to reduce the total information loss of structure and attributes, and to maintain the availability of data. Experimental results show that the method is effective in realizing algorithm function and reducing information loss.
【作者单位】：同济大学计算机科学与技术系;江西科技师范大学数学与计算机科学学院;嵌入式系统与服务计算教育部重点实验室(同济大学);江西科技学院信息工程学院;南通大学计算机科学与技术学院;
【基金】：国家自然科学基金(61762044,71561013,61402244) 江西科技师范大学重点科研项目(2016XJZD002)~~
【分类号】：TP309

【相似文献】