基于k-频繁子图聚类的二分图匿名方法  

Novel bipartite graph anonymous method based on k-frequent subgraph clustering

作　　者：吴宏伟[1,2] 张健沛[1] 杨静[1] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,哈尔滨150000 [2]哈尔滨理工大学计算中心,哈尔滨150080

出　　处：《计算机工程与应用》2013年第17期18-23,37,共7页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61073043;No.61073041);黑龙江省自然科学基金(No.F200901;No.G200827);哈尔滨市科技创新人才研究专项资金(No.2011RFXXG015);高等学校博士学科点专项科研基金(No.20112304110011;No.20122304110012)

摘　　要：针对以二分图形式发布的社会网络隐私泄露问题,提出了一种面向敏感边识别攻击的社会网络二分图匿名方法。在已有k-安全分组的理论基础上,结合二分图的稀疏性和敏感边识别攻击形式,分别提出了正单向、逆单向以及完全(c1,c2)-安全性原则,并在此基础上,形式化地定义了一类抗敏感边识别攻击的社会网络二分图安全匿名问题;同时,还提出了一种基于k-频繁子图聚类的二分图划分算法和一种基于二分图(c1,c2)-安全性的匿名算法来保证发布二分图的安全性。实验结果表明,该算法在与已有方法相当时间开销的前提下,能产生更小的信息损失度,有效地抵制了敏感边识别攻击,实现了二分图的安全发布。In this paper, a novel bipartite graph anonymous method is proposed to against sensitive edges identification attacks from malicious users and to preserve the privacy of members in social networks. A positive one-way （c1, c2）-security, a negative one-way （c1, c2）-security, and a two-way （c1, c2）-security principles are introduced. These definitions are based on the k-security group theory, the sparsity of bipartite graphs, and the sensitive edges identification attacks in social networks. A bipartite graph anonymous problem is defined to against sensitive edges identification attacks. A bigraph partitioning algorithm is presented on the basis of k-frequent subgraphs clustering and a bipartite graph anonymous algorithm is given to assure the safety of the pub- lished bipartite graph. The experimental results show that under the equal time-cost conditions, the proposed method not only produces less information loss than that of the existing methods, but also effectively resists sensitive edges identification attacks and realizes security release of bipartite graphs.

关 键 词：社会网络 隐私匿名 聚类 敏感边识别攻击 k-频繁子图 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]