Research On Anonymous Technology Of Social Network Resisting Edge Recognition In Cloud Environment

With the development of social informationization,various types of social networks have been established and used to facilitate people to make friends,interact,communicate and comment on a series of activities.Social networks require users to register and provide some personal information,including interest,occupation or income status,and even most of them need real-name authentication.The emergence and development of social network provide a good open platform for scientific research and social services to conduct various research and analysis,such as user behavior,community formation and so on.However,public research and publication of social network data bring great risks to users'personal information and security privacy.Therefore,privacy protection for social network data is particularly important and even becomes a bottleneck restricting the development of social network data analysis.Moreover,with the arrival of the era of big data and the rapid development of social networks,the number of users in social networks is also increasing.Traditional social network analysis and privacy protection technology based on the single workstation is not suitable for massive social network data.The execution efficiency and data processing ability of social network algorithm based on a single workstation cannot meet the practical application requirements.Therefore,parallel privacy protection technology has become a research hotspot;on the other hand,existing privacy protection methods cannot resist the edge re-identification attack initiated by attackers using edge structure and attribute information as background knowledge at the same time,so the intensity of privacy protection cannot meet the actual needs.Firstly,in order to improve the privacy intensity of links in social networks and the anonymity efficiency of large-scale social networks,a k-anonymity model of edge information and a distributed parallel anonymity method D-EIAM are proposed.In this method,firstly,under the distributed parallel processing system Spark,the edges of the social network are greedily grouped into groups then anonymous edge information sequences are generated.Then,based on the distributed parallel graph processing framework GraphX,anonymous graphs are generated by adding pseudo-nodes and generalizing attributes in parallel.At the same time,the attributes and structural information of links are protected anonymously to achieve the goal of privacy protection and efficient anonymous large-scale social network.Secondly,in order to satisfy the requirement of dynamic social network analysis,k~m-NMF anonymity model is proposed,in which K represents the level of privacy protection and M represents the time period when an attacker can monitor the victim.The model ensures that within the time threshold m,the probability of edge recognition in the social network is less than 1/k,anonymously protects the links in the dynamic social network to resist edge recognition attacks.Finally,the efficiency and data availability of the two anonymous algorithms are analyzed on the distributed graph data processing cluster by using the social network graph data set com-YouTube and the paper data set DBLP.The experimental results show that the proposed two anonymity protection algorithms can efficiently process large-scale social network data while ensuring the availability of published data.