Social Network-Based Trust Evaluation In Crowdsourcing Systems

In some crowdsourcing platforms based on Online Social Networks(OSNs),like Freelancer and Quora,a requester can find a group of crowd workers that can meet the requirements.During this process,untrustworthy workers can cheat the existing trust evaluation models by using some typical deceptions,like counterfeiting good reputations and overstating personal skills,to obtain fake but high trust values.Therefore,it is necessary and significant to build up an efficient and effective trust evaluation model to defense these deceptions and help deliver accurate trust evaluation results.In this thesis,we firstly propose the contextual social network structure,which has the social contexts like social trust,social relationship and social positions and the concept of Quality of Trust(QoT).Then we model the trust evaluation problem in social crowd as the problem of finding the optimal social trust path with multiple constraints of social contexts,which is the classical NP-Complete Multi-Constrained Optimal Path(MCOP)selection problem.Secondly,to deal with this challenging problem,based on the Monte Carlo method and our optimization search strategies,we propose a new efficient and effective approximation method,Context-Aware Worker Selection Algorithm C-AWSA.Besides,in order to improve the effectiveness and efficiency of our algorithm,we propose a concept of Strong Social Component(SSC)in the social networks,which emblems a group of workers who have strong connections.And we propose a novel index for SSC.Thirdly,as the contextual information in crowdsourcing can also affect the trustworthiness of the worker,we take the task based contexts,i.e.,types of tasks and reward amounts of tasks,into consideration,and we propose two classifications based on task types and task reward amount respectively.On the basis of the classifications and the workers' historical records,we propose a trust evaluation model,which consists of two types of context-aware trust: task type based trust(TaTrust)and reward amount based trust(RaTrust).Finally,we propose a novel context-aware trust evaluation algorithm CAT,which is more effective than C-AWSA,as CAT considers not only social contexts,but also crowd contexts.We demonstrate the effectiveness and availability of the proposed methods on on real-world datasets.The experimental results show that our proposed C-AWSA and CAT outperform the state-of-the-art trust evaluation methods in effectiveness and efficiency.