Based On The Social Network Privacy Protection Of Subgraph Generalization

With the development of Internet, many social networks are rise suddenly, such asMySpace, Twitter, Face-book and so on. There are more and more information whichexposed in the work. This information that exposed in the network used by attacker maybring much loss of benefits, even if it will break in the law. As a result, people areincreasingly concerned about their privacy. So the study of protecting social networkprivacy began to multiply. The native techniques of protecting data-privacy are maturity.It’s be widely used in many areas. Famous methods are anonymous, generalization,perturbations and cluster. Anonymous techniques include the K-degree anonymous, theK-neighbors anonymous, K-isomorphic anonymous and so on. Anonymous techniquesare more complicated, but it will result in less data loss. Generalization techniques andrandom perturbations are simple, but is result in more data loss. It’s very important tobalance the degree of privacy protecting and data loss.This thesis presents a method, which use the attributes of node to cluster. Thismethod is based on two technologies: K-anonymous and generalization. First, wechange the node to vector using the node’s attributes information. Second, we clusterthese vectors to some super nodes. And then, we release the statistical property of eachsuper node to public.When we talk about privacy, it always means the balance between data availabilityand the degree of privacy protection. In the thesis, we use distance-loss and degree-lossto judge the result of experiments. The experiments are based on the random networkgenerate by pajek and real networks net-science and US-air. Respectively, we study theimpact of different factors, such as the dimensions of vector, the number of cluster, thevalue of K, and so on.