The Analysis And Design Of The Privacy Protection Model Specific To The Whole Neighborhood Relational Attacks

In recent years, along with the advance of Internet technology, people are participating in more and more activities of the social network. As a result, it generates a lot of social network data, most of which contains privacy information. Due to demands, such as scientific research, social network data often tend to be released out of data publisher and the implied privacy information is under threat. Therefore, privacy protection process becomes highly important when publishing social network data.At present, there are three major problems about existing privacy protection models. Firstly, it is difficult to measure attackers’ background knowledge. Attackers can collect simple information, such as degrees of nodes, weights of edges and so on. Meanwhile it is also possible to collect complex combinations of information, such as betweenness, accessibility of paths and so on. Secondly, there are no perfect rules so far to define information loss generated by the process of the anonymity. Thirdly, anonymous methods of social networks can not be fixed. Researchers need to design their own strategies according to research questions.Researchers have made a lot of effective approaches for problems set above, such as greedy clustering method, sensitive side grouping method, weighted network community mining method and so on.The paper considers a privacy attack model based on whole neighborhood relations. Attackers can master whole neighborhood graphs of target nodes as their max background knowledge. Based on this, through analysis of the query depth, it proves that the more complexly querying, the greater probability of privacy leak.The paper proposes a SA-Weighted (security and availability, SA) privacy protection model specific to whole neighborhood relational attack, exploring and analyzing the specific protection effectiveness of different types and sizes of social network data. Three experiments, including the relationship of the probability of nodes identified and information loss, the weight distribution and the accuracy of average shortest path queries, verifies the SA-Weighted privacy model, contraposing the particular type of social network data, specifically small-scale and more closely the social network data, can not only protect its security to a certain degree, which means relatively reducing the probability of private information to be disclosed effectively, but also ensures that the released this kind of social network data has an essential availability. However, since this privacy protection model only applying to some sort of social network data, it has some limitations in the privacy protection of utility.In the era of the Internet, the research of the paper has a certain degree of theoretical and practical significance, existing some value of the anonymously publishing process of social network diagram meanwhile. To sum up, the research of the paper has certain practical value.