| With the continuous development of information technology,large amounts of data to be collected and processed in many institutions such as medical systems,social platforms,etc.Further analysis and publication of these data not only brings people many convenient,but also plays an important role in scientific research.However,the release of the data would cause disclosure of personal privacy information.Therefore,how to improve the usefulness of the released data and protect sensitive private information effectively,has an important meaning.In recent years,for the privacy concerns,a series of studies presented at the efforts of researchers.Among them,differential privacy protection model is put forward in 2006 by Dwork.This model defines a strict attack model,can provide effective protection.While providing a strict definition and quantitative assessment criteria,so that differential privacy protection method with different parameters can compare the protection level with each other effectively.Because of strong security and rigorous evaluation criteria,differential privacy model is used in various fields.Among them,the differential streaming data privacy protection has become a hot research direction.Difference privacy streaming data distribution,mainly comprising the event level of privacy protection and user level of privacy protection and so on.But in the existing method,there are still some problems.Noise accumulate over time,The algorithm efficiency is not high enough,and the strict requirements for deployment platform.There is still much research space in how to control query sensitivity,how to increase publish data availability,how to prevent privacy budget exhausted and how to improve the efficiency.The paper focus on researching differential privacy streaming data publication with non-uniform privacy budget.Aim to improve the efficiency and accuracy of query.Main contributions of the paper are as follows:(1)In order to solve the problem about noise accumulate over time and the query precision under uniform private budget is still to be improved,this paper establishes the differential privacy streaming data publication model.The model is based on sliding window,through the analysis of streaming data distribution characteristics and constructing tree structure dynamically.Based on the analysis and prediction of user query history,optimization the precision of the publication model.Theoretical analysis ancd simulation experiment results show that the algorithm is correct,and can improve the query precision effectively.(2)In the differential privacy streaming data publication model,there may be a problem of consistency constraint in the process of tree-building and noise processing.Based on sliding window and streaming data publication,under the consistent constraint,proposed an optimal linear unbiased estimation optimization algorithm.The algorithms can process real-time optimization in dynamic tree-building,ensuring high algorithm efficiency and improve the query precision effectively.Experiments and analysis show that the proposed method can improve query accuracy and efficiency.(3)Under the single sliding window,the historical data of the sliding window can not be used efficiently.Based on this problem,this paper puts forward a framework with hierarchical sliding window and adaptive capacity.The framework can effectively expand the query range,and improve practicability of the publication model.Experiments and analysis show that the framework can improve flexibility and practicality of the publishing model with high algorithm efficiency. |
