Research On Differential Privacy Protection For Task Allocation And Traffic Flow Prediction In VANET

With the rapid development of transportation informatization and wireless communication technology,vehicular ad hoc networks(VANET)have become an important part of intelligent transportation systems and have attracted widespread attention.VANET senses and collects traffic information through the vehicles and traffic infrastructures.In this way,a wide range of traffic data can be obtained at a lower cost.Traffic flow prediction,as the research hotspot in VANET,can effectively relieve urban traffic congestion.However,during the process of traffic information collection and release,there may be malicious attackers,causing vehicle’s identity or location information to be leaked and bringing privacy risks.In addition,VANET,as an application scenario of crowd sensing,also needs to pay attention to the basic problem of crowd sensing,that is,task allocation.This is of great significance to complete tasks safely and efficiently in VANET based on crowd sensing.In summary,this dissertation studies task allocation in VANET based on crowd sensing and traffic flow prediction in VANET respectively,and designs the corresponding privacy protection scheme.The main work of this dissertation is as follows:1.The privacy preserving task allocation scheme in VANET based on crowd sensing: For the scenario where the quality of service required by the service provider cannot be guaranteed due to the uneven quality of service of the users,this dissertation designs a privacy preseving task allocation scheme for users in VANET based on crowd sensing.Existing task allocation studies only consider whether the task is completed,and do not consider the quality of task completion.This dissertation screens users through data quality,and regards high-quality users as the primary choice for participating in tasks.Compared with other crowd sensing frameworks,it can ensure the high quality of task completion.Our scheme uses differential privacy technology to protect the privacy of the user’s location.At the same time,in order to reduce unnecessary privacy leakage,our scheme reduces the number of communication times between users and the platform to reduce the loss of location privacy for the entire user group.Finally,this dissertation theoretically proves the security of the scheme,and evaluates the performance of the scheme through experiments.The results show that our scheme guarantees the user’s privacy while ensuring the quality of the task.2.The privacy preserving traffic flow prediction scheme in VANET: For the scenario of urban traffic congestion,this dissertation designs a privacy-preserving traffic flow prediction scheme.Although there has been some studies on applying differential privacy into VANET,they have not considered how to reduce the error impact of differential privacy.They just carry out basic applications.Based on the characteristics of the VANET,the parallel composition of differential privacy and data correlation,a uniform grouping strategy is proposed.The privacy budget is reasonably allocated to the released dataset,which reduces the error of the noise on the released dataset and improves the availability of released dataset.In order to reduce the impact of the privacy protection mechanism on the traffic flow prediction results in the scheme,this dissertation combines exponential smoothing model with long short term memory network to make the traffic flow prediction more accurate.Finally,through theoretical analysis of the scheme,this dissertation proves the security of the scheme.Through experimental evaluation,it is proved that the scheme can provide high availability of released dataset and accurate traffic flow prediction.