Research Of The Privacy Preserving For Users In Social Networks

The rapid expansion of Internet technology has led to the emergence of various social networking platforms,and the consequent increase in the number of users on the platforms has been exponential.These social platforms provide users with the convenience of information sharing,and at the same time generate a huge amount of data which involves users’ personal sensitive information.Many scholars have currently developed data mining techniques and proposed social network analysis methods to study the value underlying these data.However,once the published data is improperly used by other malicious attackers,users in social networks may be exposed to malicious attacks,and even lead to security issues such as privacy leaks.Therefore,it is particularly important to protect the privacy of the published social network data.It is worthwhile to study how to efficiently protect user privacy while ensuring the utility of the published data.To address this issue,this thesis researches user’s privacy preserving in social networks from two aspects,that is,user’s identity privacy and user’s edge relationship privacy,respectively,and devotes to designing a method that can effectively protect user’s privacy and guarantee data utility at the same time.The details are as follows:(1)To address the problem that the publication of social network data may be subject to degree attacks,which may lead to user privacy leakage,we propose ak-degree anonymization privacy preserving scheme,which is based on the average degree of nodes.Firstly,we use the greedy algorithm based on the average degree to divide the degree sequence of social network nodes,and modify the degrees of nodes in the same grouping to the average degree,generating a k-degree anonymous sequence,which greatly reduces the distance to the original degree sequence;secondly,we modify the graph structure of the original graph using three types of edge operations: edge addition,edge deletion,and edge exchange,due to the consideration of (Neighborhood Centrality)values are taken into account when operating on edges,the important edges in the network are preserved,and the anonymized network maintains a better connectivity and relationship structure,thus improving the utility of published data;finally,simulation experiments on the social network dataset show that our proposed scheme can not only effectively improve the ability of the social network to resist degree attacks,but also maintain high stability of the network structure and high utility of published data,which can effectively protect the privacy of user identity in social networks.(2)To address the problem that the publication of weighted social network data may be attacked by edge weights and thus lead to user privacy leakage,we propose a personalized privacy preserving scheme based on differential privacy in weighted social networks.Firstly,the MCL clustering algorithm is used to cluster the weighted social network graph into different clusters with similar characteristics,in order to hierarchically process the edge weights;secondly,according to the edge weight information in each cluster,the appropriate privacy budget parameter is determined for each cluster by designing the function f(x)to personalize the appropriate differential privacy noise for each cluster,which greatly reduces the the amount of added noise;thirdly,we add Laplace noise to edge weights in different clusters individually according to the privacy budget to achieve personalized differential privacy preserving for weighted social networks;finally,simulation experiments on social network datasets show that our proposed scheme not only effectively improves the ability of weighted social networks to resist edge weight attacks,but also greatly reduces the amount of noise addition,ensuring that the perturbed network graphs still have high data utility and can effectively protect the privacy of edge relations for users in weighted social networks.(3)To address the problem that the publication of social network data may be subject to graph structure and thus lead to user privacy leakage,we propose an uncertain graph privacy protection scheme based on node similarity(UG-NS).Firstly,we calculate the similarity between nodes of the original graph using the Node2Vec model,and obtain the similarity matrix between nodes,then assign edge probability values to the edges existing in the original graph according to the similarity matrix to initially generate the uncertain graph;secondly,we add noise to the initially generated uncertain graph,add a certain number of noisy edges,and assign the generated Laplace noise as edge probability values to the noisy edges,obtaining the final uncertain graph;finally,simulation experiments on social network datasets show that our proposed scheme not only effectively improves the ability of social networks to resist graph structure attacks,but also better preserves the structural information of the original graph,providing a better trade-off between user privacy and data utility,and can effectively protect user identity privacy and user relationship privacy in social networks.