| With the development of modern computer information technology,data is becoming increasingly important as the carrier of information.Usually,the data curators will collection and summary a lot of personal data,then provide to public users by data publishing to analyze the data content,but the data will inevitably contain privacy information,so data curators need to process the data after data collected through privacypreserving model before providing to the public users to use,so as to avoid privacy leakage.Differential privacy is currently the most popular privacy-preserving model.The data publishing mechanism based on differential privacy ensures that the collected personal privacy information is not violated,and it does not affect the statistical analysis results after data publishing.To solve the problem above,The vertically heterogeneous differential privacy mechanism and the skyline query differential privacy mechanism are proposed.Vertically heterogeneous differential privacy mechanism has studied the fine-grained privacy protection problem of multiple-attributes data.Through the privacy information characterization,the vertically partitioned data is divided into different privacy information,and information entropy is used to calculate the dependence of attribute and privacy information.The privacy weights are assigned to different attributes according to the dependence,and noises are added to different attributes with the privacy weights,which solve the heterogeneous differential privacy protection problem of the vertical attributes.Skyline query differential privacy mechanism has studied the differential privacy protection of multi-objective optimization query.Through the definition of skyline query,the nondominanted utility function are proposed,and the traditional output domain generation method is improved to design a two-step noise adding strategy combining numerical and non-numerical results,which solves the scalability problem that leads differential privacy difficult to be achieved.Through experimental analysis,the above two differential privacy mechanisms not only protect data privacy and ensure certain data effectiveness in different application scenarios.Under a fixed privacy budget,the vertical heterogeneous differential privacy mechanism achieves the goal of achieving different levels of privacy protection for different attribute properties.The results show that under the proposed mechanism,the higher the privacy protection level,the better the trade-off level between privacy gain and data effectiveness,and is better than the homogeneous differential mechanism.The skyline query differential privacy mechanism also achieves the differential privacy protection for multiobjective optimized queries.The results show that the mechanism is useful,and the more irregular the data distribution,the more data and the more dimensions,the lower privacy protection level,then the better the trade-off level of privacy gain and data effectiveness.In addition,the proposed mechanism has good fault tolerance. |
