Research On Privacy Protection Scheme Of Social Data Publishing

In recent years,with the rapid development of Internet applications such as Weibo,and the popularity of personal mobile devices,social network has developed and become an integral part of people's daily lives.Social network provides people with an online platform for sharing information on interests and hobbies,as well as huge commercial value.However,the popularity of social network and the release of social network data have also led to the leakage of user's privacy information.Therefore,how to protect the privacy of users and ensure the utility of data while data publishing has become a hot research topic.This thesis deeply researches the data of social network,divides the data into graph data and tabular data,and proposes corresponding privacy protection methods for both types of data.For the graph data,this thesis considers the sparsity of each column vector in the graph matrix,and applies compressive sensing technology to the privacy protection of social network for the first time.We use compressive sensing technology to randomly perturb the graph structure to protect the user's privacy information.For the tabular data,this thesis first considers the four attacker models in the tabular data,designs the corresponding privacy model,and proposes a privacy protection method based on Anatomization.Anatomization includes two processes,table division and group division.During the table division,we first adopt entropy and mean-square contingency coefficient to partition attributes into separate tables to inject uncertainly for reconstructing the original table.During the group division,all the records in the original table are partitioned into non-overlapping groups so that the published data satisfies the privacy requirements of privacy model.This thesis separately carries out corresponding experiment for the two privacy protection methods based on real data sets.At the same time,detailed privacy analysis and utility analysis are given for each method in this thesis.Experimental results show that the two privacy protection methods can effectively protect the privacy while ensuring high utility of the data.