Research On Trust Model And Trust-based Application For Grid Computing

Along with emergence of grid technology, the user management isn't central, closed and controllable. Grid system is facing all kinds of threats that include threats from the inner of system and threats from the outer of system. Traditional computer security mainly researches on preventing intrusion from the outer of system. However, traditional method has nothing to do with threads from the inner of system. So, it is necessary to evaluate the network behavior of users in grid computing. Thus, evaluation of the user's behavior can reflect his reliability. This dissertation studies how to build trust model for grid computing in order to find out malicious users and improve security and reliability of system.The contributions of this dissertation are as follows:1) A trust model that takes domain as unit and has two layers is presented for grid computing. This trust model definitely gives definition of trust for grid computing, and analyzes that trust has attributes of asymmetry, transitivity and dynamics. This trsut model satisfies requirements of trsut in grid environment. This trust model describes behavior trust and identity trust in detail and formalizes representation of trust relationship network. The trust model is low complexity of computing, scalable and autonomous.2) A trust envaluation model that is one of submodels of trust model is presented.This trust envaluation model describes the direct trust relationship of domains and envaluation of member in a domain. In order to resolve the problem of direct trust relationship, this dissertation develops the Bayesian-Theory-Based Direct Trust Evaluation algorithm (BTBDTE algorithm). The characters of BTBDTE algorithm are as follows: when the trust value is higher, the trust value increase slowly; In the case of the same trust value, when evaluation of the transaction is negative, the range of trust value decreasing is bigger than that of trust value increasing when transaction is positive. BTBDTE algorithm follows the rule of trust value changing. But BTBDTE algorithm has the drawback that can't instantly reflect the current vibration of entity's reliability. So, an Improved BTBDTE algorithm was developed. The IBTBDTE algorithm not only has characters of BTBDTE algorithm, but also can instantly reflect the current reliability of entity. In order to resolve the problem of member envaluation in one domain, a method that compute the trust value of the member in one...