Research On Privacy Protection Mechanism Of Mobile Crowdsourcing In Edge Cloud Environment

With the rapid development of Mobile Internet,5G network and smart mobile device technology,Mobile Crowdsourcing(MCS)technology has been widely used in the real world,greatly facilitating people’s lives.However,while MCS brings convenience to people,people also face the risk of privacy leakage that cannot be ignored.In addition,with the increasing amount of interactive data in MCS,traditional cloud computing has been difficult to meet the high real-time and processing efficiency of data.The proposal and application of edge computing technology provides a new solution to this problem.However,edge computing usually requires edge servers to complete computing services.As a semi-trusted role,edge computing servers may still leak the privacy of crowd workers.Most of the existing MCS complete the data interaction based on the trusted third platform and complete the data processing through the server side,without considering the threat of third-party platforms and servers to the privacy of crowd workers.According to the spatio-temporal characteristics and task demand characteristics of MCS,the acceptance and submission of tasks by crowd workers will expose the location and track information of workers,and the submitted task data will also involve sensitive information of workers.Therefore,the privacy of crowd workers is mainly divided into location trajectory privacy and task data privacy.This thesis focuses on the protection of two kinds of privacy in MCS under the edge cloud environment,proposes different privacy protection mechanisms to realize the security protection of various private information of crowd workers.The main research content of this thesis is as follows.1.Aiming at the problem of location trajectory privacy leakage,this thesis proposes a triple real-time location trajectory privacy protection mechanism based on edge computing and blockchain.The local differential privacy algorithm is used to prevent external attackers and third-party roles from directly obtaining the location and trajectory information of crowd workers.Using the characteristics of Gaussian distribution,combined with the current location status and privacy sensitivity of crowd workers,a multiple probability extension algorithm is designed to probabilistically extend the sensitive locations in the trajectory and increase the difficulty of identifying sensitive locations in the trajectory.Utilizing the dynamic pseudonym algorithm and the spatio-temporal characteristics of privacy,a spatio-temporal dynamic pseudonym algorithm is designed to reduce the correlation between crowd workers and trajectories through the random irrelevance of dynamic pseudonyms.Distributed distribution tasks and data uploads are realized through the blockchain,avoiding untrustworthy third-party platforms from participating in task interactions,ensuring the privacy and security of crowd workers.2.Aiming at the problem of task data privacy leakage,this thesis designs a verifiable federated learning privacy protection mechanism based on edge computing and blockchain.Using edge computing technology to complete data near-end computing and improve system operation efficiency.Combined with blockchain technology,data transmission is completed through broadcasting and decentralized crowdsourcing task interaction is realized.Utilize the framework of the federated learning algorithm,complete the distribution and submission of task requirements through model interaction and ensure that the data of crowd workers does not go out of the local area.Based on the random response mechanism and local differential privacy algorithm,a double local disturbance local differential privacy algorithm is designed to complete the localization processing of crowd workers’ data privacy and ensure data privacy security.Through the historical task scoring of crowd workers,a reputation calculation algorithm is designed,and the set of crowd workers is selected using the reputation value to ensure the quality of crowd workers.Combining prediction ideas with federated learning algorithms,a verifiable federated learning algorithm framework is constructed to screen the local model data of crowd workers to ensure overall service quality.