Research On Data Quality Assurance Of Crowdsensing In The Internet Of Things

With the development of mobile intelligent terminal devices and the advances of Internet of Things(Io T)technologies,an increasing number of user devices with robust sensing and computing capacities have been engaged in data sensing and aggregation,after that,the sensed massive data are uploaded to diverse sensing platforms for data-centric services.The abovementioned phenomenon has given rise to a crowdsensing-based novel data sensing and computing paradigm for Io T.In this paradigm,the guarantee of data quality is complicated when considering the high location sensitivity,large scale,and profit-seeking nature of participating users.A lack of secure,efficient,and low-cost data quality assurance methods will inevitably reduce the enthusiasm and sensed utility of participating users,and hinder the development of Io T applications,which are challenging to achieve the security,efficiency,and rationality requirements of crowdsensing-based computing paradigm.To solve this problem,this dissertation studies the data quality assurance of crowdsensing in the Internet of Things and focuses on the issue of critical technologies,especially in data security-preserving,data selection,and data revenue allocation,which are respectively used to solve a series of problems of user location data security preservation,data selection and matching,clustering user evaluation,and reasonable revenue allocation respectively,the main content involving:(1)To solve the problem that the current location security-preserving methods rely on the trusted sensing platform and ignore the differences in user scale and security requirements in various regions,a location security-preserving method based on user scale is proposed.This method leverages the edge servers to design a user location security-preserving framework,avoiding dependence on a trusted sensing platform and realizing distributed user location security-preserving.At the same time,according to the user scale in different regions,this method proposes location security-preserving algorithms in popular regions and remote regions,respectively.For popular regions,the traditional k-anonymity method is improved to satisfy the high-efficiency requirements of location security-preserving in large-scale users.For remote regions,data encryption is adopted to meet the high-security needs of location security-preserving in small-scale users.Finally,the feasibility and effectiveness of this method are verified by comparing it to other traditional and effective security-preserving methods.(2)Aiming at the problem that the data redundancy of large-scale users and the selection preference of task publishers in data selection are ignored,a user data selection method based on selection preference is proposed,which considers the influence of data quality evaluation and sensing cost selection.Firstly,the traditional clustering method is improved by defining user similarity and setting user scale,which aims to balance user distribution,reduce data redundancy,and improve the accuracy of high-quality user aggregation.Meanwhile,considering the preferences of task publishers,the sensing cost selection is transformed into a multi attribute decision making problem.The personalized selection of user data is realized by the prospect theory and the VIKOR method.Finally,the user clustering accuracy and selection diversity of this method are verified by simulation experiments.(3)In view of the problem that the accurate value is used to evaluate the data quality of clustering users in data selection while ignoring the internal user differences and resulting in evaluation errors,a data selection method based on clustering is proposed.Firstly,the union-initiators are used to calculate the similarity and form a sensing user-union.The collected user-union data is transformed into interval-valued data,which aims to reflect the differences of internal users accurately.Then,this method utilizes relative entropy and introduces preference coefficient to realize the quality evaluation of clustered user data and the selection of differentiated user data.Finally,the effects of internal user differences and selection preferences on the accuracy of this method are verified by simulation experiments.(4)A revenue allocation method based on user learning ability is proposed to solve the problem that the existing user revenue allocation methods ignore the identification of abnormal users and learning ability.Firstly,the publisher-user evolutionary game model is constructed for the revenue allocation problem by improving the traditional evolutionary game model and considering the user learning ability.Then,the error elimination decision theory is used to realize the identification of abnormal users.Through the evolutionary stability strategy solution and stability analysis,the optimal user revenue selection strategy under different conditions is obtained to realize the user revenue allocation.Finally,the abnormal user identification accuracy and the revenue allocation rationality of this method are verified by simulation experiments.