Research On Trust Mechanism-Based Task Scheduling In Ad Hoc Networks

In the military, field exploration, space exploration and other fields, the communication technologies are quite different from those used in the traditional network. They should not rely on the network infrastructure but be self-organized into a network. Ad hoc network rightly meets the requirement, but because of the characteristics, such as no center point, dynamic topology, limited resources, and so on, it is more vulnerable to be attacked than traditional wired network, and also traditional security technologies can not be directly applied to ad hoc network, so the study of ad hoc network security and trust management is of great significance.This dissertation discusses the security threats with which the ad hoc networks face, the concepts and characteristics of trust, and different kinds of trust models for ad hoc networks. Through the discussion, we find that traditional PKI-based security policy can not meet the security requirement of the current dynamic and uncertain ad doc environment, and trust mechanism becomes a effective supplementary for ad hoc network security problem. In this paper, considering the demand for lower limit of trust and the impact of the importance to trust values, we set up a trust model considering the trust demands of different tasks, and then we apply it to the ad hoc task scheduling. At the beginning, based on historical records and recommended information of other nodes, the scheduler evaluates the trust degrees of resource nodes, and then assigns the tasks to the resource nodes which can meet their demands for the lower limit of trust. After completing these tasks, the results and lower limits of trust demand of the tasks will be recorded. This dissertation uses the trust demand to measure the importance of tasks, and applies it into the trust model to influence the evaluation of trust, so the task can be allocated to the nodes which meet the demand. At the same time, it can increase the probability of success and reduce the total completion time, the performance is improved and the network is well operated. This dissertation first analyzes the classical Min-Min algorithm, and then based on the consideration of trust and resource limit, an extended Min-Min(TD-Min-Min) task scheduling algorithm is proposed. From the experiment, the results show that, at the minimum completion time and the number of failure tasks, the performance of TD-Min-Min algorithm is better than Min-Min algorithm.In real life, people usually observe various aspects of a person to determine whether he or she can be trusted. Same as above, in the complex network the nodes also need various factors to judge the credibility of other nodes. Considering the impact of multi-factors to trust evaluation, we set up a multi-factor trust model. In this model, the node gets multi-factor information from some trusted nodes, but in some cases, some trusted nodes use accumulated trust to provide malicious information, so this dissertation uses the outlier detection technology to shield malicious information, and improves the reliability of recommended information, which consequently assures the accurary of trust judgements.