Abstract: With the rapid development of social networks,social networks have accumulated a large amount of data,which reflect the social laws to some extent.Aiming at mining effective knowledge under the premise of ensuring privacy,this paper proposed a clustering-based social network privacy protection method.The method has the characteristics of adaptive privacy protection strength,high security and effectiveness of anonymity model.Clustering is conducted based on user information and social relationships.It clusters all nodes in the social network into a super point containing at least k nodes according to the distance between nodes,and then the super points are anonymized.Anonymous super points can effectively prevent various types of privacy attacks taking node attribute privacy and sub-graph structure as background knowledge,so that attackers cannot identify users with a probability greater than 1/k.According to the characteristics of clustering algorithm and social network,the initial node selection algorithm and node spacing calculation method in clustering process are optimized,and by combining the adaptive thinking,the selection method of privacy protection strength is also optimized,which effectively reduces information loss and improves data validity.Experiments were carried out on Matlab platform with different data sets.The results show that the proposed method issuperior to other related methods in terms of information loss and running time,which further proves its effectiveness and security.

Key words: Clustering, Information security, K-anonymity, Privacy protection, Social network