Cloud-edge Collaborative Privacy Protection Mechanism Based On Differential Privacy

With the rapid development of 5G and the Internet of Things,edge computing plays a more and more important role in real-world applications.Edge nodes provide users with rich personalized services by collecting a large amount of user data which may lead to leakages of privacy.There is a problem of effectively estimating data with multi-level privacy requirements in cloud computing,since users have different privacy protection requirements.It's a research hotspot to ensure user's personalized privacy requirements and make full use of data for effective estimation with multi-level privacy requirements in cloud-edge collaboration scenarios.The paper first introduces the current privacy protection technologies and attack problems,and then focuses on the problems of the mean estimation and histogram estimation algorithms in the differential privacy protection model: the traditional 1-Bit differential privacy does not specify the influence of the selection of the probability interval on the privacy protection effect.The reduction in data volume will cause a larger error of mean estimation since 1-Bit differential privacy is a pure differential privacy.The traditional personalized local differential privacy based on data derivation does not consider the influence of coding on the error of histogram estimation,and the data derivation algorithm has a high algorithmic complexity.The paper solves the above problems in the cloud-edge collaboration scenario,and the main work is as follows:To solve various problems in the data derivation(Data Recycle with Personalized Privacy,DRPP)algorithm,an OUE-ODRPP algorithm based on Optimized Unary Encoding(OUE)is proposed.The main work of this paper is as follows:(1)The paper analyzed the effect of probability interval and interval length on mean estimation in the 1-Bit differential privacy model,proposed and proved three instructive inferences: the impact of whether the probability interval is symmetrical on the privacy level;the influence of the offset distance of the probability interval on the privacy level;the effect of probability interval length on privacy level and estimation error.(2)In order to optimize the accuracy of mean estimation and the flexibility of the 1-Bit differential privacy,the paper proposed a relaxed 1-Bit differential privacy protection model based on the inference results,which using a small amount of privacy budget to greatly increase the stability and estimation accuracy of the model.(3)The paper proposed a distributed and personalized local differential privacy protection model under the cloud-edge collaboration scenario.To optimizes the algorithmic complexity of the traditional DRPP(Data Recycle with Personalized Privacy)algorithm and quality of histogram estimation,the paper proposed the OUEODRPP(Optimized Data Recycle with Personalized Privacy based on Optimized Unary Encoding)algorithm based on OUE(Optimizes Unary Encoding).Then the simulation and comparison of the proposed model,algorithm and inferences are carried out.The simulation results verified the correctness of the theory and optimization performance in the paper.Finally,the paper summarized the research results and looked forward to the future of cloud-edge collaboration privacy protection technology.In this paper,we use 16 figures,9 tables,51 references.