Research On De-anonymous Social Network Attack Based On Node Multi-hop Features

With the rapid development of online social networks, the demand of publishingand sharing social network data for the purpose of commercial or research isincreasing. However, the disclosure risks of sensitive information of social networkusers are also arising, so how to protect the user’s sensitive information has becomethe focus of people’s attention. Early use of simple anonymous method to protectprivacy, with the depth study of privacy protection, at present privacy protectionmethods are: k-anonymous methods, cluster-based on the generalization method,graph randomization method. These privacy protection methods have an effectiveprotection on user’s privacy, but there are still some limitations. When the attacker toattack privacy according to their own different types of background knowledge, therehave different types of structural attacks. However, in all of this, the auxiliary graphas background knowledge to anonymous attack can be used in the real network dataon the attack.Based on the graph mining research technology and the de-anonymous socialnetwork attacks, the paper proposed a structural attack based on the node n-hopneighbor feature n-hop NeighFNR. The method is using the node n-hop neighborfeature acquisition node from the multi-dimensional characteristics of nodeinformation, starting from the anonymous graph structure, using the auxiliary graphbackground knowledge, implementation of a structured attack method to identifynodes in the graph of the anonymous. The nodes on the matching process, using theweighted graph matching algorithm based on simulated annealing, we can get the bestmatching result between the auxiliary graph and anonymous graph, so as to achievethe purpose to identify anonymous graph node. Set karate and email on the test in thereal social network data show that: this method relies only on the attack of networkstructure can effectively in simple anonymous graph and randomization anonymousgraph to identify nodes. For Karate this small data set can be used to identify the datanode, for large data sets can be used as seed nodes identification in the de-anonymousattack. This study provides a reference for the social network graph data attack andprivacy protection.