A Probabilistic Trust Model for Handling Inaccurate Reputation Sources

Download

	PDF Download (436Kb)
	Postscript Download (682Kb)

Description/Abstract

This research aims to develop a model of trust and reputation that will ensure good interactions amongst software agents, in large scale open systems in ral and in Grid systems in particular. In this context, we identify the following as key drivers for our model. Firstly, the Grid is an open environment in hich an agent cannot be aware of all of its interaction partners. Furthermore, there is a possibility that these interaction partners may be self-interested and provide false accounts of their experiences with other agents (either by lying or by having an inaccurate world view). Secondly, the openness and m of Grids means that agents will need to interact with other agents with which they have no past experience. Nevertheless, even in this context, such must try to assess the trustworthiness of their interaction partner. Against this background, we developed TRAVOS (a Trust and Reputation model for Agent-sed Virtual OrganisationS) which focuses on providing a measure of trust for an agent to place in an interaction partner. This measure of trust is d by considering the past experiences between the agent and its interaction partner, and using these to form a probability density function. In instances n there is no personal experience, the model substitutes past experience with reputation information gathered from other agents in the society. In this r case, we pay particular attention to how our model provides mechanisms that can be used to handle the possibility that this reputation information may be naccurate, due to poor modelling or malicious intent.