Research On Key Techniques For Mutual Trust Mechanism In Grid

Towards the two main ambitions of wide area resource sharing and cooperative problem-solving, Grid computing has emerged as one of the key computing paradigms that enable the creation and management of Internet-based utility computing infrastructure for realization of e-Science and e-Business at the global level. Fusing with SOA (Service Oriented Architecture), Grid applications are becoming increasingly richer. But, to bring Grid computing into large scale deployment and an effective support for global scale sharing and cooperation, trust is the single largest issue to be overcome. In traditional closed environment and those distributed environment with small scale resource sharing, for the relative familiarity between users and the limited number of entities, trust relationship in cyberspace is usually directly mapped from physical trust relationship existing in real life, the effect of trust is obscure; but for the inter-domain, open and dynamic Grid environment, there is prevalent strangeness between entities, the gurantee from trust cannot be overlooked. There is an urgent need for a new mutual-trust mechanism for Grid, which is adaptive to the characteristics of Grid environment and with scalability, flexibility, dynamic adaptability and robustness in a whole.In this paper, we focus on service Grid, root the whole work in the definition of trust and its basic characteristics, clarify the relationship between trust and reputation, trust and security, analyze the current progress in credential-based and reputation-based trust systems, show their advantages and disadvantages, and then from the speciality of trust establishment in Grid, study several key techniques in Grid mutual trust systems, which include: mutual-trust infrastructure, trustworthiness evaluation based on credential, reputation evidence analysis and aggregation in reputation-based trust mechanisms, federated evaluation and decision-making with multiple trustworthiness result involved. Our main contribution can be summarized as follows:(1) Research on mutual-trust infrastructure in GridWith the introduction of trustworthiness into Grid, we propose a trustworthiness-aware contract-supervised Grid mutual trust infrastrucuture TaCs (Trustworthiness-aware Contract-supervised), which aims at promoting reliable and flexible mutual trust relationship establishment, leverage the predictability of cooperation, and boost benign cycle in Grid. In the evaluation of trustworthiness, we integrate the two mainstream mechanisms: credential-based and reputation-based trust systems, which makes a good supplement between each other. TaCs sufficiently fuse three elements: credential, reputation and contract. The introduction of reputation aims at bridging entities, relating transactions, and sharing experiences; the introduction of credential aims at satisfying the need from highly trusted computing and making up for trustworthiness evalution when reputation evidence is few or even none; to improve operability and controllability, TaCs is based on the state-of-the-art service oriented architecture, combined with SNAP (Service Negotiation and Acquisition Protocol) [113], and forms a contract-supervised mechanism, which clarifies the context of trust and avoids dishonest feedback flooding the whole environment.(2) Improvement on credential-based trust mechanism in GridFor the uncertainties and differentiation with credential in Grid, we propose the concept of reputation trustworthiness factor. And with ideas from uncertainty reasoning, we give methods for trustworthiness evaluation with single credential involved and multiple credentials involved.(3) Research on key techniques for reputation-based trust mechanism in GridIn chorus with the specific characteristics of Grid environment, we study the two crucial components of reputation-based trust mechanisms: reputation evidence analysis and aggregation. For the subjectivity of trust and the prevalent strangeness between entities in Grid, we propose a pre-evaluating set-based bias-tuning method, with which reputation evidence from different entities are endowed the ability to be compared and aggregated; For the two kinds of dishonest feedback: collusive and individual dishonest feedback, we propose effective filtering methods for them respectively, which is an enhancement to the robustness of our reputation mechanism; For the diversity of evidence sets in Grid, we divide the evidence aggregation problem and tackle them separately according to the different characteristics of evidence in amount, time distribution and rater origin. Namely, we propose three aggregation methods: the entity behavior characteristics based reputation aggregation method, which is suitable for evidence in large amount, with broader time span and richer rater origin; the rating criteria similarity based reputation aggregation method, which is suitable for evidence in little amount, with clustered time distribution and limited rater origin; the grey prediction theory based trustworthiness interval predition method, which is suitable for evidence in limited amount, with a certain time span and even time interval. Simulation results and comparison between similar work in Grid show the effectiveness of these methods.(4) Research on trustworthy decision-makingFrom the multi-faced characteristic of trust, we study federated evalution and decision-making with multiple trustworthiness status involved, and propose two methods: the fuzzy partial order based method with no target given and the grey correlation decision-making based method with a reference target given, which provide operable guide with good overalbility for real practices such as resource selection, authorization decision and access control deployment in Grid.