Research On Key Technology Of Privacy Preserving For Network Security

Human has entered an era of rapid development of information technology,network data also largely increases.Behind the data there is a large number of more important information.Through the further analysis of data,finding the relationship and rules between the data can bring huge benefits.So data mining technology has been in great development.However,people's privacy is likely to leak while mining data.In response to this information content security of network security,privacy-preserving data publishing technology has been proposed and earned many researchers' concern.K-anonymity is the basic technology of privacy-preserving data publishing.Because in the same equivalence class there must be K records at least,ensuring one record in equivalence class indistinguishable with other K-1 records.So the probability of identity was confirmed don't exceed 1/K.It can effectively prevent the linking attack.Based on the global generalization of Kanonymity model is a common model,but its loss of information is relatively large.This paper proposes an algorithm of K-anonymity?L-diversity based on clustering.We try to choose similar data for a cluster which satisfies K-member.Re-clustering those do not meet L-diversity constraints in order to resist the homogeneous attack.Experimental results show that the algorithm can effectively resist homogeneous attack and reduce the information loss of data sheet.However,it does not consider the record's sensitive attribute.If the records belong to a class of sensitive or high sensitive attribute properties in the same equivalence class,the data sheet may be subject to a similar attack or high sensitive property speculation attacks.So we pay close attention to record's sensitive properties,proposing an anonymity clustering algorithm based on sensitive attributes distribution.The data table is re-sorted in accordance with the distribution of reorder's sensitive attributes.Firstly we choose the record that has high sensitivity or belongs to a class.Here sensitive attributes coverage concept is presented.Only satisfying the records' coverage,the record can be inserted into the cluster.One record is inserted the cluster at each time.We firstly consider record's sensitive attribute privacy protection other than information loss.Experimental results show that although its information loss is more than K-anonymity,L-diversity algorithm based on clustering.It can truly withstand similar attack or high sensitive property speculation attacks and has less information loss than K-anonymity model based on global generalization.