Research On Key Technologies Of Crowd Computing For Mobile Social Networks

With the popularization of mobile devices and the rapid development of wireless communication technologies,crowd computing is becoming a newly emerging distributed problem solving model under the“Internet+”background,which makes full use of the built-in rich sensors in handheld smart devices(GPS,acceleration,cameras,gyroscopes,etc.)and increasingly powerful storage and computing capabilities,as well as users' intimacy and movement patterns in social networks,replaces traditional large-scale static sensor networks to solve problems that are difficult for machines or individuals to accomplish.Crowd computing has the advantages of easy deployment,flexibility,resource saving,etc.,so it has broad application prospects in the real world and is also an important part of the development of smart cities.Starting with the key steps of the crowd computing system,the crowd computing online incentive mechanism,location privacy protection and real-time task assignment in the mobile social network are mainly studied in this thesis.The corresponding improved algorithms are designed to ensure that the crowd computing system can run efficiently and with high quality in relevant scenarios.The main contents of this thesis are as follows:1.The current situation of research on crowd computing at home and abroad and the crossbackground with other disciplines are fully investigated,problems and difficulties in the current research are summed up to determine the direction of this thesis.Then,the theoretical basis of crowd computing is introduced,including the basic principles of origin and application,system structure,and key steps.The basic algorithms of privacy protection and its application in crowd computing and the basic knowledges of mobile social networking are also introduced,to lay the foundation for the follow-up research.2.A crowd computing user recruitment and incentive mechanism based on reputation control is proposed.For the user-initiated crowd computing scenario represented by urban wifi signal monitoring,attention is paid to the online real-time user recruitment and incentive problems in this scenario,aiming at maximizing system efficiency under certain budget and time constraints.In the online scene,users randomly arrive at and leave the mission area,the locationbased computational task coverage and perceived contribution degree are fully considered to model,and an improved multi-stage reverse auction algorithm is proposed firstly,which adaptively adjusts density threshold at each stage through online learning and dynamically selects the optimal user set.After the end of each transaction,the reputation of the user is updated according to the quality of the task completion,and punishment is imposed on the party who breached the contract.The real-time reputation value carried by the user is instead introduced into the benefit function definition of the above reverse auction algorithm to guide the user recruitment in the next stage.Theoretical analysis and simulation results show that the incentive mechanism proposed in this thesis satisfies the four basic principles of computational validity,personal rationality,platform profitability,and authenticity,and can obtain better value benefits under certain time and budget constraints.3.A crowd computing task assignment algorithm based on differential privacy protection is proposed.For the platform-initiated crowd computing task scenario represented by the space environment monitoring,the privacy protection strategy is added during the process of the platform acquiring the users' location and performing the global task assignments.A third-party trusted institution is introduced and different privacy space decomposition(PSD)algorithms are adopted to scramble real-time location information of users,and then the scrambled statistical results are sent to the crowdsourcing platform,reasonable greedy algorithms are designed to perform task assignments.Based on the success rate of task allocation,the optimal balance between task allocation efficiency and privacy protection effect is achieved.The simulation results of both real and simulated data demonstrate that the task assignment algorithm based on differential privacy protection can effectively protect users' privacy while only losing a small task allocation efficiency.Different differential privacy policies also have different advantages and disadvantages.For example,the task assignment algorithm based on the contour line PSD policy is relatively insensitive to the privacy budget,so is more suitable for applications with higher privacy protection requirements.4.A crowd computing online task assignment algorithm based on meeting prediction in mobile social networks is proposed.Taking into account the crowd computing model carried on mobile social networks,the intimacy and encounter laws between users are used to publish tasks and transport data back,and large-scale data transmission is achieved through near-field communication.In this scenario,a timeline parallel model of task execution and user encounter is proposed,at the same time,different users' calculation ability differences of tasks are considered,two improved online task assignment algorithms are designed to minimize average task feedback time and minimize longest task feedback time separately.Simulation results show that the improved model is more in line with the multi-task,multi-heterogeneous users crowd computing scenarios where data requirements are increasing rapidly.It can achieve better time gains for the same task and user size,and improve task completion efficiency,so has a strong practical and promotional value in the actual crowd computing scene.