PG-TRUST: A SELF-ADAPTIVE AND SCALABLE TRUST COMPUTING MODEL FOR LARGE-SCALE PEER-TO-PEER GRID COMPUTING

Abstract

Peer-to-peer Grid (PG) is a new, promising approach for distributed computation that takes advantage of the huge amount of resources in grid computing. The purpose of PG is to support cooperative work between large-scale grid domains. In order to protect each domain's privilege and security, an efficient trust management mechanism is essential. Focusing on the requirement of large-scale PG computing, PG-Trust, a novel trust model is proposed based on trust relationships across domains. Firstly, a self-adaptive overall trust quantified framework is proposed based on the historical interaction records between participants, which not only can improve system efficiency, but also can solve trust computing problem when the direct evidence is sparse. Furthermore, the direct trust computing method based on attenuation function and the feedback trust mechanism based on direct trust tree (DTT) are introduced, which makes our model have a higher self-adaptability and a better scalability than previous studies. Through performance analyzing, the proposed model exhibits a strong self-adaptability in handling various dynamic behaviors of peers, and has a remarkable enhancement in the accuracy of trust evaluation.