Research On Privacy Preserving Techniques Based On K-Anonymity For Data Publishing In The Social Network

With the vigorous development of network information technology,Web2.0 network has went deeply into people's life. The frequency for use of social websites such as Facebook, Twitter, Sina Weibo and We Chat, has significantly increased. For the purpose of data analysis, social network data are collected and issued with high volumes, however, this process is always along with disclosure of privacy information. Therefore, the contradiction between the public serviceability of social network and privacy information protection among internet has been a focus of attention during scholars.On the basis of that the usability of effectively issuing data should be assured, this paper researched into multiple aspects of the privacy protection technologies of the data publishing in social network.Firstly, for the sake of resisting against the recognition attack from attackers based on background knowledge of social network graph structure, a model of privacy preserving for social network data issuance is established in this paper. The model can effectively combine the advantages of different models, using different methods in different parts of the data sets, thus improves the level of privacy preserving, and provides ideological guidance for future research.Secondly, in order to solve the problem of significant data loss and poor availability in a single mode of social network privacy preserving, a privacy preserving method based on k- isomorphism and locally randomized is proposed.It is divided into three steps:(1) The graph information is simply anonymized;(2) The k- isomorphism method is optimized, the graph is divided effectively into subgraphs and those subgraphs are processed to be isomorphic, and then,the hash coding method would be used to determine whether the subgraphs meet the requirements of isomorphism anonymity;(3) The randomization method is optimized, and some subgraphs are processed under restriction conditions,though which the graph radius is restricted within a certain range, network graph can be eventually issued with anonymity. Theoretical analysis shows that the proposed method satisfies k-isomorphic definition, and network graphs issued would meet the k- security requirements, meanwhile, network structure information has also been well maintained, the validity of the user data can be guaranteed, and it can also effectively resist recognition attack from attacker based on background knowledge of graph structure.Finally, according to test results, the privacy preserving method based on kisomorphism and locally randomized proposed in this paper can effectively reduce information loss, and has good performance in measurements of harmonic mean of the shortest distance and subgraph centrality of information within graph. It can protect user's privacy information effectively and improve the usability of data issued.