Research On Trust Model And Trust-based Access Control For Grid

Known as the next generation Internet, grid computing integrates various geographical distributed computing resources, heterogeneous software and hardware into the networks to construct a virtual supercomputer to solve the problems of large-scale applications. Thus the security of grids outstands as the grid is featured with geographical distribution, software and hardware heterogeneousness, dynamic and multi-domain. Specialists of grid security have proposed the resolutions of subjective trust model after they observed that human conduct intercourse on the basis of trust. That is to evaluate the subjective trust degree based on the previous transaction result of participants of grid, and then to address the security issues occurred in grid transactions.However, with the popularity and increasing scales of grid, the traditional grid subjective trust model exposed unsolved problems. Those problems are: the difficulty for grid service providers to evaluate the trust degree efficiently and precisely for very large number of users; the difficulty to prevent the whitewashing attacks as the malicious users keep changing ID to take attacks; the difficulty for grid users to evaluate the trust degree of grid service providers when the number of transaction is less enough, etc. Besides, there are defects originating from recommendation trust processing method of traditional trust model, such as: that the grid entities hardly find recommender when acquiring recommendation trust of target entity; that the length of the trust chain or any malicious recommender in the trust chain would seriously effect the accuracy of the trust; that the traditional recommendation trust processing method do not distinguish between the trust of recommending and the trust of transaction.At the same time, the traditional access control method, such as the access control list (ACL) and role-based access control (RBAC), decide whether to permit access by judging the certification and credential of users be in accordance with the policy. The traditional methods obviously could not meet the needs of security raised by grid service provider because they have not considered the behavior and performance of users in the previous transactions. In addition, though the majority of subjective trust-based access control methods have taken users'previous transaction into consideration, they could not avoid users cheating in more intelligent way because that the access control decisions are made up only upon the satisfaction of a threshold value of user's trust.This thesis has proposed corresponding solutions to solve the above-mentioned problems, with the main content and novelties illustrated as follows:1) A user trust agent-based subjective trust model framework has been proposed. Within this framework, each autonomous domain sets up a user trust agent to manage the trust degree of all users in this domain. The grid service provider only needs to judge the honesty of trust agent by observing transaction results and the transaction users'domain trust degree. A simple trust model and relevant experiments have proved that the trust model designed from this framework could be able to precisely solve such problems as elvaluating large amount of user trust degree efficiently, prevention of whitewashing attacks and difficulty for grid user to make precise evaluation on trust degree of grid service provider in circumstance of limited number of transaction.2) A cloud model-based recommendation trust processing method has been put forward. By this method, evaluator acquires the recommendation trust of evaluatee by means of broadcast, which free him from worrying about length of trust chain and malicious recommender existing in trust chain. After that, the evaluator takes use of cloud model to form recommendation trust cloud, and then decide evaluatee's recommendation trust degree by comparing the clouds. Experiments showed that this method could be able to kick off many defects of recommendation trust processing method existing in traditional trust model.3) A cloud model-based grid subjective trust model has been designed. This model integrated the advantages of above-mentioned trust model framework and recommendation trust processing method in a perfect way. In this model, the grid service provider evaluates the user trust agent for different user's domain trust level, which is evaluated by the user trust agent, calculates the agent's direct trust value according to the result of transactions with corresponding trust level users, and acquires the agent's recommendation trust for the corresponding trust level by means of broadcast, and finally takes use of cloud model to form trust cloud for corresponding trust level in order to judge the authenticity of evaluation on corresponding trust level user given by the user trust agent. Meanwhile, the user trust agent takes use of cloud model to evaluate trust level of grid service provider to give reference for domain user's evaluation on the grid service provider's trust. Experiments showed that this model, in accord with grid architecture and the subjective trust's fuzzy and random, made the mutual trust evaluation between grid service provider and grid user more accurate and rational.4) A access control policy for grid service provider based on game theory and trust has been proposed. Regular grid access control methods use digital certificate and credential to verify the authenticity of users, ignore the behavior of users, can not protect against the security threats from interior. Most of the trust-based access control could only examine whether the user trust degree meets the threshold value, that gives the malicious users the opportunity to obtain additional profit by means of controlling the number of attacks. The access control policy proposed in this thesis applies the attribute in game theory to help forward both grid user and service provider to toe the mark when making transaction, so as to earn benefit to the utmost from successful deals. Experiments showed that this method could curb attempts of malicious attacks of grid users effectively and encourage grid users and service providers to try their utmost to make transaction successful.