抗复合攻击的社会网络(k,l)匿名方法

(k,l)-Anonymity for Social Networks Publication Against Composite Attacks

针对社会网络发布时由于复合攻击所带来的隐私泄露问题,提出了一种(k,l)-匿名发布隐私保护方法.首先在k-同构和l-多样性的理论基础上,给出了复合攻击形式和图的(k,l)-匿名模型,并形式化地定义了一类节点具有单敏感属性的简单无向图的(k,l)-匿名问题.同时,提出了一种基于k-匿名和l-多样性的属性泛化算法来解决该匿名问题.实验结果表明:该算法能产生比已有方法更小的信息损失度,以及相当的时间开销,可有效抵御复合攻击,保护发布社会网络的隐私信息.

Privacy preserving in social networks has raised serious concerns in recent years. One of the privacy disclosures is brought about through composite attacks（ structural attacks and attributes attacks） by malicious users. In this paper, we consider the privacy preserving publication against composite attacks in social networks, which are expressed by simple undirected graphs. Firstly, we present a （k,l）-anonymity graph model based on the theoretical principle of k-isomorphism and the l-diversity, and then a （k,l）-anonymity problem faced the simple undirected graphs is formally defined. In addition, we propose a generalization algorithm based on k-anonymity and l-diversity to solve the problem defined above. The experimental results show that under the equal conditions, our method not only produces less information loss than that of the departed method, but also needs less time cost, which effectively resists composite attacks and preserves privacy information of the published networks.