Research On High Quality Participant Selection And Data Assessment Methods In Mobile Crowd Sensing

Recently,more and more people have begun to pay attention to the Internet of Things(IoT).This model has provided great convenience for people’s lives,and has aroused widespread research interest,one of which is Mobile crowd sensing(MCS).It is a new data acquisition mode that combines crowdsourcing ideas and sensing ability of multiple mobile devices and uses people to sense and collect data according to specific requirements.MCS has the advantages of low cost,convenient maintenance,good scalability,and it has been applied to air monitoring,intelligent transportation,urban management,and other aspects.In MCS,the quality of selecting participants directly affects the quality of submitted data,while data quality assessment techniques directly affect the enthusiasm of participants.Both have a decisive impact on the final data results of sensing tasks,becoming two important research directions of current researchers’ attention.In addition,with the increasing awareness of users’ security,privacy and security issues during data collection and submission have become one of the bottlenecks in the development of MCS and have attracted the attention of researchers.Although there have been corresponding programs to solve the above problems,the following problems still exist: 1)In the process of participant selection,the sensing quality of participants is inferred mainly through historical record analysis and model prediction.However,due to the timeliness of task historical data and the dynamic nature of MCS environment,the sensing quality of each participant cannot be effectively calculated;2)The privacy security protection measures in MCS are mainly encryption and anonymity technology,but these partial methods have the risk of high load sensing platform and insufficient privacy protection effect in MCS;3)The existing data quality assessment techniques are mainly designed from the historical records of participants and model detection.But the low accuracy of data assessment and the presence of abnormal data items in this part of the technology have not significantly improved the quality of the ultimately sensing data.Therefore,in order to solve the above problems,this thesis puts forward the following two solutions,the main contents are summarized as follows:(1)This thesis proposes a blockchain-based reputation sharing high quality participant selection method(SAPS)to address the issues of insufficient reliability and privacy leakage risks in existing MCS participant selection schemes.By using participant reputation data from multiple domains,this method can select reliable participants and ensure the security of participant privacy data.Firstly,this method proposes a blockchain-based framework for trusted participant selection of reputation sharing.The framework uses nodes of blockchain to replace centralized sensing platform in MCS to achieve decentralization,and effectively obtains the reputation of participants by integrating reputation data from multiple domains.Meanwhile,the framework can ensure the credibility of reputation data by using third-party trusted platforms;Secondly,SAPS designs a blockchain-based incentive mechanism to reward blockchain nodes that act as sub-sensing platforms;Finally,the privacy of participants’ reputation data is protected by distributed data storage and differential privacy.Experiments show that SAPS outperforms the comparison methods in both dense and sparse data sets.In particular,participants in SAPS have 6% lower selection cost and16% improvement in task completion quality compared to the comparison methods.(2)This thesis proposes a brainstorming-based data quality assessment mechanism(BSDA)to address the issues of insufficient accuracy and low sensing data quality caused by abnormal data items in existing MCS data quality assessment methods.Firstly,BSDA achieves decentralization by replacing the sensing platform in MCS with nodes on the blockchain;Secondly,BSDA utilizes brainstorming to analyze the data quality of sensing tasks between nodes(sub-sensing platforms).Through multi-angle assessment,BSDA avoids the problems of low data accuracy and abnormal items in the data,and records the assessment criteria by each node after obtaining reliable data results;Finally,BSDA periodically updates the data assessment standards obtained through brainstorming to ensure the reliability and accuracy of subsequent data quality assessment.Experiments show that the performance of BSDA is superior to the comparison methods on different data sets.Specifically,the data quality index of BSDA has improved by 7% compared to the comparison methods.