Research On Privacy-Preserved Task Distribution Algorithms For Mobile Crowd Sensing

Mobile Crowd Sensing(MCS)is a novel,flexible and efficient data sensing approach,which collects sensing data needed by the task requester with a group of users who carry mobile intelligent devices.Benefiting from the powerful sensing,computing and commu-nication capabilities of mobile sensing devices and the low cost of crowdsourcing,MCS has been widely used in environmental sensing,infrastructure sensing and social sensing.Taking map collection for an example,Baidu Map,Gaode Map and Waze Navigation and other platforms distribute the map collection tasks to the eligible participants.Participants will receive corresponding rewards from the platform after reporting road conditions such as congestion,construction,accidents,etc.Currently,one of the core difficult problems in MCS,is that how to distribute tasks so that more sensing tasks can be completed at a lower cost.However,when collectors participate in the map collection and contribute the sens-ing data,their privacy information face the risk of disclosure,e.g.,geographical location,movement route,personal identity,etc.Concerning about privacy disclosure,the partic-ipants' enthusiasm and the quality of sensing tasks will decline dramatically.Therefore,protecting the participants' privacy is an urgent problem to be solved in task distribution for MCS.Currently,a variety of privacy-preserved task distribution methods have been pro-posed for MCS.However,they are still imperfect and have some defects.First,they rely on the trusted platform.The trusted platform is responsible for protecting participants' privacy from being known by the third party.In fact,a fully trusted platform is non-existent in real applications.Second,researchers mainly focus on protecting the single-location privacy,while research on path privacy protection is insufficient.This can not meet the path privacy protection needs of participants who perform a set of tasks sequentially.Third,the platform provides undifferentiated privacy protection for different participants,which can not satisfy the various demands of participants.Focusing on the privacy-preserved task distribution problem in MCS,this dissertation proposes some privacy-preserved task distribution algorithms that do not rely on a trusted platform to solve the above problems,and provide the personalized privacy protection for participants in different application scenarios.The main contents of this dissertation are as follows:(1)Location-level path privacy-preserved task selection.Aiming at protecting par-ticipants' location privacy during participating in sensing tasks,this dissertation proposes a location privacy-preserved task selection algorithm.For the participant,the expected benefits of participants are maximized under the constrained mobile distance and location privacy protection requirements.The location differential privacy mechanism is adopted to protect each passing location on the path,and the partial publishing method is used to protect the destination location of the path against Bayesian inference attack.Experimen-tal tests validate that the proposed algorithm can provide participants with location-level path privacy protection and obtain high task revenue.(2)Trajectory-level path privacy-preserved task selection.In order to protect participants' trajectory privacy during taking sensing tasks,this dissertation proposes a trajectory privacy-preserved task selection algorithm.For the participant,the expected benefits of participants are maximized under the limited mobile distance while providing trajectory privacy protection.Firstly,a novel trajectory privacy is defined according to the distance from the trajectory to the set of sensitive locations.Secondly,an optimized distributed trajectory privacy-preserved task acquisition algorithm based on dynamic programming is proposed.Finally,a centralized trajectory privacy-preserved task assignment algorithm relied on a trusted platform is proposed to evaluate the cost of not relying on trusted plat-form.Experimental tests validate that the proposed algorithm can provide trajectory-level path privacy protection and achieve high task completion rate.(3)Location privacy-preserved reverse auction task distribution.For preventing location privacy disclosure from bid tasks and bid price in reverse auction,this dissertation proposes a location privacy-preserved task distribution algorithm based on reverse auc-tion.First of all,the inference attack model is established from the perspective of the attacker,and the participant's task bidding is used to infer the source location of the participant.Secondly,a privacy-awareness task selection algorithm is proposed for each participant.Finally,a privacy-awareness task assignment algorithm is provided for the platform.Experimental tests validate that the proposed algorithm can protect the partici-pants' location privacy without increasing the data acquisition cost of the platform.