A Research On Privacy Preserving Technology For Dynamic Graph Data Publication

With the rapid development of mobile internet technology,online social networks have become an important part of modern life and have a profound impact on human social behavior.Analyzing the data in the social networks can help us understand the topology of the network,reveal the evolution of social networks,and it is also of great significance to stock forecasting,disease tracking and public opinion analysis.However,the data contains a lot of personal privacy.If these data are simply processed and directly used in data mining research,it will often lead to the leakage of user's privacy.So how to protect the privacy of published data has become an important research area.Existing privacy protection technologies focus on single data releases,which means that no any change is made after the data is released.However,taking into account the dynamic evolution of social networks,there is a strong correlation between published data by different time nodes,so the dynamic characteristics of social networks determine the privacy preserving method of a single data release can not fully guarantee dynamic social network privacy and security.Therefore,how to meet the sequential data release requirements of dynamic social networks and avoid inference attacks caused by different published data has become an important research hot spot in the field of privacy protection.Next is a brief introduction to our main work.The first work,aiming at the problems of high time complexity and large amount of cumulative information loss in the process of sequential data release,we propose an incremental graph anonymity framework based on restart of anonymous process.The framework uses activated function to determine when the anonymous process needs to be restarted,which ensures that the anonymous strategy can be adaptively selected to achieve the desired performance of anonymization.Then based on our proposed incremental anonymity framework,we abstracts the social network into a simple graph,and for the degree-trail attack that exists in the dynamic publishing scenario,we propose an efficient incremental simple graph anonymous algorithm.From the attribute characteristics of the simple graph,we verify that our algorithm provides high data utility while ensuring privacy.It also verifies that the anonymity framework reduces the accumulated information loss after the anonymous process restarts,and improves the anonymous data utility.The second work,also based on our proposed incremental anonymity framework proposed in first work.We abstracts the social network into a weighted graph,and proposes an attack model of weight-bag trail attack in the dynamic weighted graph publishing scenario.Based on the attack model,we propose an efficient incremental weighted graph anonymous algorithm.From the attribute statistics of the weighted graph,we verify that our algorithm provides high data utility while ensuring privacy.It also proves that our anonymity framework achieves a good balance between privacy and data utility.