Research On Differential Privacy Protection Based On Related Attribute

Because of the rapid development of Internet technology,the public has also changed the corresponding way of life and production.In the past few decades,a network of daily living has created an Internet epoch.Mobile network equipment have gradually become more and more increasing using of our daily life.Simultaneously,Internet companies compete with each other,they provide users with quality and free services,so that they can obtain a huge number of users and data information resources.Due to frequent occurrences of user privacy leaking incidents,individuals and businesses are also increasingly concerned about privacy issues.This thesis takes differential privacy protection as the main research topic,and studies the differential privacy protection method for related attribute link attacks.In the first part,this paper summarizes and summarizes the classic privacy protection issues and its existing research results,while summarizes the differential privacy protection technology in data release.In the second part,some basic principles and methods of the differential privacy application are described.In the third part,we studied the utility and measure of correlation between two main attributes of data set,designed a differential privacy protection LDAA algorithm for the correlation of quasi-identifier attributes and sensitive attributes,so that the technology can ensure the security of user data privacy information,and will not directly affect the actual availability of shared data,and achieved the main purpose of this research.Using the experimental design of the datasets,the results show that when a privacy protection model of the quasi-identifier attribute is associated with the sensitive value attribute meets certain specific conditions,the-diversity model can effectively prevent link attack,meet the time complexity of the simplified algorithm and greatly improve the availability of data,which combines good availability and privacy,thus realizing the technology of data release for privacy protection of personal information when the attributes are associated with the owners.This paper mainly studies the differential privacy protection method against related attribute link attacks Under the condition of valid data release.This article summarizes the existing research results in the field of privacy protection and differential privacy,and explains the most basic principles and characteristics.On the basis of effective data release,the ?-differential privacy and the refinement of k-anonymity are two popular privacy protection models.However,the two privacy protection modes mentioned here inevitably have drawbacks.In order to take advantage of their advantages and avoid their disadvantages,we propose an(l,?)-diversity model to implement privacy protection data publishing.Our method uses sampling to make the data to be published first meet the?-differential privacy,and then we use mutual information to find the most accurate quasi-identifier attributes associated with sensitive attributes before anonymizing them.After that,we use the generalization method to make the published data set satisfy the(l,?)-diversity model.The experimental results can plenary show that our privacy protection model has good creativity and excellent running ability,and the algorithm has very good efficiency.Figure 10 Table 3 Reference 50...