MODELING TRUST IN OPEN DISTRIBUTED MULTIAGENT SYSTEMS

摘要

Many software applications are open distributed systems whose com-ponents are decentralized, constantly changed, and spread throughout network. Peer-to-peer networks, web service, semantic web, e-business, m-commerce, autonomic and pervasive computing are among such systems. The need for autonomy and flexibility in interaction of components has been becoming essential for their efficient activities. These systems can be modeled as open multi-systems that are composed of autonomous agents interacting with .each other according to mechanisms and proto-cols. In the recent years, a number of models of interaction among agents such as coordination, collaboration and negotiation have been proposed. However, the applications of these ineraction models in large-scale open distributed sytems face with new challenges because of openness and self-interest of agents. These agents have their own goals and may join and leave at any time. The problems are how agents decide who and when to interact with each other. That means agents need to deal with degrees of uncertainty in making decisions. In order to minimise the uncertainty during interaction in open distributed systems, agents have to trust each other. Trust is considered to be a belief that an agent has with a coun-terpart on holiest, reliability or reciprocativeness for the common goal. The paper is to consider models of trust to enable agents to calculate the amount of trust they can place on their interaction patners. to make reasoning of propogation of trust among agents during interaction. In this way, models may guide agents in making decision on who, how and when to interact.