Research On Reputation Mechanism In Internet-based Virtual Computing Environment

Internet-based virtual computing environment (iVCE) aims to provide powerful, flexible, and convenient computing ability and harmoniously unified, secure, and transparent services to Internet applications by aggregating and integrally utilizing of widespread resources. Finally, realize the change of internet from"potential computing platform"to"trustworthy computing platform". However, because internet resources have dynamic and autonomic characters, the research and development of iVCE confront with much new challenge. In which, from the angle of trustworthy, behavior of autonomic element (AE) in iVCE is uncontrollable and rational that iVCE confronts with much issues as free-riding, unreliable services, strategic speculation in service and tendentious speculation in rating, etc. The flooding of such problems will directly induce the debasing of trustworthiness in system, and service quality will be hard to ensure. Therefore, it is necessary to establish effective, perfect reputation mechanism to restrain and guide AE's behavior, and then system's trustworthy and harmonious development would be promoted through the upgrade of AE's behavior trustworthiness. It has been an important idea and urgent needs in internet computing and in the research and development of iVCE.Around the key question of constructing incentive-compatible reputation mechanism in iVCE, bases on the thoroughly analysis for current research on reputation and incentive mechanism, this paper puts forward researches on the key techniques in aspects: penalty-incentive mechanism based on reputation, method for evaluating evidence of reputation, incentive mechanism for credible rating, and application-oriented trustworthy resource selection method. Compared with existing researches on reputation systems, this paper makes many evident progresses from reasonableness of reputation evaluation, effectiveness and self-enforcing of incentive, ability of resisting misbehavior, judging credibility on rating and providing method for service selection. The main contribution of this paper is detailed as follow:1) In view of the present situation that the service availability is seriously affected by misbehavior of selfish and malicious peers, to encourage AEs'active participation and honest collaboration, the paper presents a reputation-based penalty-incentive mechanism, named PEtrust. PETrust bases on repeated game theory to analyze AE's internal profit change in the process of trading, introduces reputation evaluating and distinguishing mechanism in reputation recovery, defines triad description of behavior's trustworthy recognition, reputation, and behavior's deviation recognition and correlative calculation among them. Thus, to embody the correlative changes among reputation, penalty-incentive, and AE's profits, and encourage AE collaborate in order to obtain maximum benefits. AE's deviation will cause reducing in reputation, the penalty period is decided by AE's contribution, reputation status, time decay factor and random probability, and penalty can be dynamically adjusted with fine granularity. The distributed implementation of PETrust is presented which bases on anonymous storage of reputation information. Theoretic analysis and simulation results show that, PETrust can distinguish and evaluate reputation effectively according to different feature of behavior, encourage AEs'collaboration enthusiasm effectively, improve system's entire efficiency, and provide better capacity of resisting collusive deception. PETrust is incentive compatible with the design goals of our reputation system. Furthermore, PETrust presents both low time complexity and few incurred packets, which is favorable for engineering deployment and implementation.2) To judge AE's credibility on rating, a D-S evidence theory based method, named mechanism for evaluating evidence of reputation (MEER), is presented. At first, calculates consistency measurement basing on rank-correlative method of kendall's concordance coefficient, and combines with similarity measurement in distance, the basic probability assignment function (BPAF) on each feature item is constructed. So probability assignment on identify elements can be got. Then, uses Dempster combination rule to aggregate all probability, gets AE's current credibility on rating. The time complexity of MEER is analyzed. Adaptive mechanism and safety enhancing method are introduced to improve the efficiency and dynamic adaptability. Case analysis and simulation experiment show that, MEER has high recognition rate on dishonest rating, low misjudgment rate, and evaluating ability for directional evidence.3) To encourage AE rate actively and honestly, the paper presents an incentive mechanism for credible rating, named IMCR. Trustworthy degree on service and trustworthy degree on rating as double indicators for measuring AE's reputation quality are introduced. IMCR calculates AE's trustworthy degree on rating basing on honest degree and participation degree on rating, and calculates AE's trustworthy degree on service basing on consumer's honest rating degree, trustworthy degree on rating, and current rating about satisfaction. By associating double indicators dynamically, realize differential service mechanism, which encourage AE rate honestly in order to obtain more high quality service. The distributed implementation of IMCR is presented. Abundance experiments show that: IMCR can depict AE's changing status of trustworthy degree for different behavior feature effectively, realize differential service basing on different rating trustworthiness, has better incentive effect compared with other similar model. IMCR is simple and effective, which is convenient to be deployed in distributed environment. And it is applicable to erecting reputation system that is incentive compatible with the goal of the whole system's trustworthiness.4) With multiple-facets of reputation status and the preference when selecting service considered, a reputation-based method for trustworthy service selecting is presented. Analyzing and concluding multiple reputation facets and selecting preference,using decision-making techniques in the theory of fuzzy set, making information aggregation and forming selection order of optional resources as reference.