加权社交网络敏感边的差分隐私保护研究

社交网络边权重表示节点属性相似性时,针对边权重能导致节点敏感属性泄露的问题,提出一种利用差分隐私保护模型的扰动策略进行边权重保护。首先根据社交网络构建属性相似图和非属性相似图,同时建立差分隐私保护算法;然后对属性相似图及非属性相似图边权重进行扰动时,设计扰动方案,并按扰动方案对属性相似图及非属性相似图进行扰动。实现了攻击者无法根据扰动后边权重判断节点属性相似性,从而防止节点敏感属性泄漏,而且该方法能够抵御攻击者拥有最大背景知识的攻击。从理论上证明了算法的可行性,并通过实验验证了算法的可行性及有效性。

Research on Differential Privacy Protection of Sensitive Edge of Weighted Social Network

When the edge weights of the network represent the similarity of the nodes" attributes, the edge weight can cause the sensitive attributes of the nodes to be leaked, and a strategy based on the differential privacy preserving model is proposed to protect the edge weights. Firstly, the attribute similarity graph and non-attribute similarity graph are constructed according to the social network, and the differential privacy protection algorithm is established. Then, the perturbation scheme is designed to make the attribute similarity graph and the non-attribute similarity graph perturbation perturbation scheme. The attacker can not judge the similarity of the attributes of nodes according to the edge weight, thus preventing the leakage of sensitive attributes of nodes. And,this method is able to withstand an attacker with the greatest background knowledge of the attack. The feasibility of the algorithm is proved theoretically, and the correctness and validity of the algorithm are verified by experiments.