Research On Privacy-preserving In Network Data Publishing Based On Anonymity Technology

At present, most privacy-preserving algorithms take the privacy requirements as the same in social network data publishing, which leads to excessive privacy-preserving to some users, ignores user’s original intention of sharing information, and reduces the availability of social network data; These algorithms are mainly based on the static social networks, so attackers can access to user’s privacy information by observing the change process of social networks; In some social networks, users also have sensitive attributes, such as salary and community; What’s more, keeping sensitive edge weights may better reflect the characteristics of social networks, and reduce information loss.In view of the above problems, the research aims to protect the privacy information in social networks by using the anonymity technology. Firstly, recent studies on social network privacypreserving are summarized in the paper, including the type s of social network privacy information, attack ways, the principle of anonymity, anonymity methods and some popular anonymity models. After analyzing the existing methods of privacy-preserving about the three models for social networks, the paper applies the personalized anonymity strategy to privacy-preserving, and designs a personalized privacy-preserving method based on dynamic social networks and a personalized privacy-preserving method based on weighted social networks.In dynamic social network data publishing, the privacy requirements are divided into three levels, and the paper improves(k, △d)-anonymity algorithm and k-neighborhood isomorphism algorithm. The method also prevents the sensitive attributes and sensitive edges disclosure. Experiments show that, compared with the(k, △d)-anonymity algorithm and k-neighborhood isomorphism algorithm, personalized anonymity algorithm is more efficient, and causes less information loss. Although sometimes it is more volatile, the overall trend is still superior to the other two algorithms.In weighed social network data publishing, the privacy requirements are also divided into three levels. By using k-degree anonymity for grouping and modifying the weight bags, the weight bag in each group meets k-anonymity and the sensitive attribute in each group meets l-diversity. Experiments show that, compared with the k-histogram- inverse-l-diversity anonymity algorithm and k-histogram anonymity algorithm, personalized anonymity algorithm is more efficient, causes less information loss. With the increase of k value, the superiority of the personalized anonymity algorithm is more ob vious. With the increase of l value, the execution time increases, and the growth rate is also increased.Whether it is applied in dynamic social networks or in weighted social networks, compared with existing algorithms, personalized anonymity algorithm not only achieves the user’s privacy requirements and improves the efficiency of implementation, but also reduces the information loss of the original social networks and improves the validity of data.