A Data Release Method For Social Network Privacy Protection

With the development of the industry of network applications, the social network, as anew kind of interpersonal communication platform, has got tremendous development. For thepurposes of research or publicity, the operational sides of the social networking sites need torelease the data held by them to the public. The loss of privacy takes a potential risk ofsecurity in the process of data release. To avoid leakage of personnel sensitive information,the data should be taken privacy protection before release.There are many scholars who dedicated to the study of social network privacy protection,and have yielded a lot of meaningful results. Although existing achievements are efficient insolving the problem of the privacy protection on personnel attributes or on communitystructure, but not taking into account all two aspects. To address this problem, we research adata release method for the social network privacy protection in this paper. The main works ofthis paper include:First of all, we propose a kind of uncertainty generalization strategy for personnelattributes. The space of possible worlds is expanded through the uncertainty generalization.This generalization strategy not only expands the space of possible worlds limittedly, but alsomakes the data which is generalized carrying the statistical information of original data.Secondly, we propose a kind of disturbing method to the community structure. Thismethod disturbs the local community structure through three means, i.e.(1) exchanging thenodes with same degree,(2) inserting nodes selected from similar community, and (3)deleting a part of friends relations which are connected to the nodes inserted. This method notonly has the ability to induce the attacker looking for wrong targets, but also can make thedisturbing not enlargement while disturbing the local community structure.Third, we put forward a privacy protection parallel algorithm, called k-N-anonymous, forsocial network, which is based both on attribute uncertainty generalization strategy and oncommunity structure disturbance method. On the one hand, the describing information ofperson meets the requirements of k-anonymous in a same community. On the other hand, the number of similar communities is not less than N in a network. Otherwise, the effectivenessof the k-N-anonymous algorithm is verified through lots of experiments.In a word, the data release method for the social network privacy protection is reasonable.This method takes more significant advantages than traditional methods in terms of theprivacy protection both on personnel attributes and on community structure.