Research And Implementation Of Truth Discovery Under Local Differential Privacy

In the face of data conflicts in multi-source data,the truth discovery task can estimate the reliability of information sources and discover the data truth.It has a wide range of applications in areas such as crowd sensing and crowdsourcing.However,when users participate in the truth discovery task,directly uploading original data which contains personal sensitive information will lead to personal privacy leakage.Therefore,how to effectively complete the truth discovery task while protecting the users'personal privacy information has become a key issue that needs to be solved urgently in the truth discovery task.Local Differential Privacy Technology(LDP)does not require a trusted third-party server.By allowing users to locally disturb data before uploading data,only users themselves can access the original data,which fundamentally prevents the leakage of users' personal privacy.Therefore,this thesis applies LDP technology to the truth discovery task to protect users' personal privacy information.The existing truth discovery work that satisfies LDP is only suitable for independent privacy protection of the single data.It is not suitable for simultaneous location disturbance and data disturbance in the truth discovery task of spatial events,or continuous data disturbance in the dynamic truth discovery task.Therefore,this thesis designs and implements a truth discovery algorithm of spatial events and a dynamic truth discovery algorithm that satisfy LDP.Aiming at the truth discovery task of spatial events for privacy protection,this thesis proposes a truth discovery algorithm of spatial events based on optimal location sampling that satisfies LDP.In this algorithm,in order to solve the problem that a large number of locations lead to the large disturbance error,this thesis proposes an optimal location sampling method based on information entropy;in order to solve the problem of low data utility caused by location disturbance and data disturbance independently,this thesis proposes a joint perturbation method in location and data of associated perception.The results of experimental tests on real data sets show that the truth discovery algorithm of spatial events that satisfies LDP proposed in this thesis significantly improves the MAE change index when compared with existing work,which verifies the superiority of the proposed algorithm.Aiming at the dynamic truth discovery task for privacy protection,this thesis proposes a dynamic truth discovery algorithm based on voting mechanism that satisfies LDP.In this algorithm,in order to solve the problem that a large number of time points lead to the large disturbance error,this thesis proposes a time point sampling method based on voting mechanism;in order to solve the problem of low accuracy caused by only using the current time point's data to estimate the users'weight,this thesis proposes a user weight estimation method based on incremental update.The results of experimental tests on real data sets show that the dynamic truth discovery algorithm that satisfies LDP proposed in this thesis significantly improves the MAE value index when compared with existing work,which verifies the superiority of the proposed algorithm.