Research On Privacy-preserving Truth Discovery In Mobile Crowdsensing

Mobile crowdsensing is one of the most influential emerging technologies,which leverages the existing communication infrastructure(Wi Fi,4G/5G)and mobility of smart device users to collect personal and surrounding environment,location,traffic conditions,noise level and other data from smart devices scattered in the monitoring area,and then store it in the cloud server to extract and deliver valuable information.Mobile crowdsensing has broad application prospects in various aspects of smart cities such as environmental monitoring,intelligent transportation,service recommendation and indoor positioning.However,the widespread application of mobile crowdsensing still faces many challenges.First of all,the sensing data submitted by users usu-ally involves sensitive personal information.If it is not protected,user privacy may be leaked.Second,due to the heterogeneity of mobile devices and the instability of user sensing behavior,the sensing data submitted by users is not always reliable or even invalid.Such heterogeneous data seriously damages with the quality of system data services.To solve these problems,it is necessary not only to protect the privacy of users’ personal data,but also to conduct truth discovery on the collected data to deduce the most reliable data results,so as to ensure the qual-ity of system data service.Therefore,how to improve the reliability of truth discovery while protecting user data privacy is a core problem of mobile crowdsensing data service system.In recent years,many researchers have proposed the privacy-preserving truth discovery mechanism for mobile crowdsensing to solve this core problem,but the existing work still has the following problems: First,for the data flow of continuous sensing tasks,there is no privacy protection mechanism that supports dynamic users and is easy to implement.Second,for the impact of user-submitted outliers on the reliability of truth discovery,there is a lack of efficient privacy protection mechanism that can guarantee the reliability of truth discovery.Third,for the impact of sensing participants on the reliability of truth discovery results,there is a lack of effictive method that supports both user dynamics and privacy and guarantees the reliability of truth discovery with less total payment.In order to solve the problems and technical challenges in existing research work,this dissertation aims at protecting user data privacy and improving system data service quality,and combines the dynamics of sensing participants and the limited resources of mobile devices.And the following explorations are carried out from different aspects that affect the reliability of true discovery results in mobile crowdsensing:(1)For the streaming characteristics of mobile crowdsensing data and the characteristics of users exiting the sensing activity at any time,this dissertation proposes a real-time privacy-preserving truth discovery on crowdsensed data stream.First,a secure summation aggregation protocol is proposed for the high dynamics of sensing participants.The protocol completely solves the problem of incorrect calculation results caused by users exiting at any time during sensing activities,and its single-server architecture is easier to deploy and implement.Sec-ondly,a dynamic incremental truth discovery algorithm is designed to significantly reduce the computing cost and communication overhead of privacy-preserving true discovery.Finally,based on the secure aggregation protocol and the dynamic incremental truth discovery algo-rithm,a privacy-preserving real-time truth discovery scheme is constructed on data stream.It not only protects the privacy of user data,but also ensures the correctness of truth discovery results.Theoretical analysis verifies system security and fault tolerance,and the experimental evaluation results show the efficiency of the system.(2)For the impact of outliers submitted by users on the reliability of truth discovery,this dis-sertation proposes a reliable and efficient privacy-preserving truth discovery mechanism.First,an encryption data filtering algorithm is proposed to enable the sensing platform to accurately identify and remove outliers from the encrypted sensing data,so as to ensure that the truth discov-ery algorithm can run well to obtain reliable results.Then,in order to improve the efficiency of the system,an inner product function encryption algorithm is proposed.Based on this algorithm,a privacy protection truth discovery mechanism is constructed to reduce the privacy-preserving truth calculation from the traditional two rounds of additive aggregation to one round of inner product.In addition,a function decryption key generation algorithm supporting double-layer encryption is proposed to protect the confidentiality of the results of true discovery.Theoreti-cal analysis and experimental evaluation prove the security,effectivness and efficiency of the scheme proposed in this dissertation.(3)For the impact of sensing participants on the reliability of truth discovery,this disser-tation proposes a dynamic pricing mechanism to protect privacy and support robust sensing.First,a privacy-preserving data aggregation algorithm is proposed to ensure the data privacy of users in sensing activities.Then,by using the aggregation algorithm,a privacy-preserving data quality evaluation method is constructed to protect user data privacy while obtaining a dynamic user quality profile in real time.Finally,a privacy-preserving dynamic pricing mechanism is constructed based on reinforcement learning.Under the condition that no knowledge such as user perception mode is available and no content of user sensing data is disclosed,this mecha-nism can adaptively learn and release optimized price of sensing data according to the quality profile of user sensing data,stimulate high quality sensing while restrain low quality sensing,realize the balance between minimum total payment and robustness of the system,and ensure that the truth discovery algorithm runs on a high-quality sensing data set to yield reliable re-sults.Theoretical analysis and experimental evaluation prove the security and effectivness of the proposed scheme.The dissertation deeply analyzes the characteristics of sensing data and users in mobile crowdsensing,and proposes corresponding privacy protection solutions in response to the ex-isting research problems.The proposed schemes are verified by theoretical analysis and ex-perimental evaluation.Compared with existing schemes,the proposed solutions have better practicality and higher performance while protecting user’s data privacy.It provides support for the construction of secure and reliable mobile crowdsensing data service system,which has certain theoretical research significance and practical value.