Research On Differential Privacy Modeladn Data Protection In Cloud Computing

With the continuous application of big data technology in the various walks of life,the development of social informatization and networking results in the explosive growth of data,and the total global data doubles approximately every two years,which brings about a new challenge for information storage,processing and security.More and more users put services on the cloud and store data in cloud computing data center.However,because users need to access the data stored in the cloud many times,or to use large-scale data for data analysis,data mining and other requirements,which makes how to ensure the security of data stored in the cloud and the data privacy is not disclosed in the process of data publishing become the main problems facing the development of cloud computing.This thesis mainly studies the technology of data privacy protection based on the differential privacy model in cloud computing.According to the study and discussion of the characteristics of the differential privacy model and the data publishing based on the differential privacy model,this thesis proposes a method to select the relevant privacy parameters in the differential privacy model and the data publishing algorithm to satisfy differential privacy model.The main contents of the study are as follows:Firstly,this thesis briefly introduces the concept,privacy security of cloud computing and the related concepts of privacy protection model,and summarizes the domestic and international research status of cloud computing,the security of cloud computing and the privacy protection of cloud computing.At the same time,the thesis also introduces the classical algorithm of privacy protection model.Secondly,in differential privacy data protection,the most important process is to add appropriate noise values to the data set in order to disrupt the data set.By analyzing the probability density function and the probability distribution function image model of distribution model,it is concluded that the selection of probability value of different noise values in the image of probability distribution function corresponds to thedifferent distribution intervals in the image of probability density function.Therefore,we propose three types of noise selection intervals.For each type of interval,we establish a function and function image model for the probability that the noise value corresponds to the different privacy parameter ? falls within the interval.In the end,we propose the upper bounds of privacy parameter ? under different noise selection intervals.Thirdly,on the basis of the traditional differential privacy histogram publishing algorithm,Boost algorithm,and in the process of transforming histogram into tree structure,this paper proposes the idea of constructing query tree by using line segment tree.The Boost algorithm needs to convert the histogram into the form of a full d-cross interval tree,however,it is often difficult to meet this requirement in practical applications.The idea of converting histogram into line segment tree is proposed in this paper.The noise that satisfies the Laplacian distribution is added to each node in the transformed line segment tree,and the weight of the node is optimized by the optimal linear unbiased estimation to meet consistency constraints between nodes.In addition,the effectiveness of the algorithm is proved by experimental evaluation.