Trust and Security in Grid Computing Environments

This project is aimed at establishing secure trusted relationships in Grid computing environments. Typically, Grid computing takes place in uncertain environments with casual relationships based on the “good name” of entities at the moment of collaboration. At the base of every decision for an interaction between the two parties, trust considerations should reside.

Trust is a complex issue in a Grid, because there are many uncertainties with respect to the level of reliability that the participants should assign to each other and it is not possible to know in advance whether a certain entity can be trusted or not. Trust values can be accumulated and calculated based on past direct reciprocal interactions or indirect interactions. Trust relates to belief in honesty, truthfulness, competence, reliability, etc., of the trusted person or service. Socially inspired trust models are useful for Grid environments, especially for the ease of efficiency in supporting the management of trust requirements. The bigger the level of accumulated trust of a participant, the better will be its “social value” and “position” in the system.

Our approach is to employ Bayesian networks for managing trust in an uncertain Grid environment, motivated by the fact that Bayesian networks allow decisions to be taken based on the accumulated and processed experience. Another idea is to consider every single entity in the environment. Each of the entities should be able to decide on their own regarding to their assessment policies. Thus, an important goal is to deal with the “decentralization” and “generalization” of the notion of trust.

For each of the participants, their identity and their behaviour are considered to establish trust among them. When trusting an entity, it is important to know which aspect we are referring to, because there are instances where one entity is trusted more than the others regarding to different levels of trust. Trust towards a participant is handled in different contexts. These contexts are used to decide whether a participant is eligible for a certain activity and the overall value of trust of a participant interferes with the decision of improving its “social position” among the others.