Year: 2012
Type: dissertation
Abstract: Multi-Agent systems (MAS) are artificial societies populated with \ndistributed autonomous agents that are intelligent and rational. \nThese self-independent agents are capable of independent decision \nmaking towards their predefined goals. These goals might be common \nbetween agents or unique for an agent. Agents may cooperate with \none another to facilitate their progresses. One of the fundamental \nchallenges in such settings is that agents do not have a full \nknowledge over the environment and regarding their decision making \nprocesses, they might need to request other agents for a piece of \ninformation or service. The crucial issues are then how to rely on \nthe information provided by other agents, how to consider the \ncollected data, and how to select appropriate agents to ask for \nthe required information. There are some proposals addressing how \nan agent can rely on other agents and how an agent can compute the \noverall opinion about a particular agent. In this context, the \ntrust value reflects the extent to which agents can rely on other \nagents and the reputation value represents public opinion about a \nparticular agent. Existing approaches for reliable information \npropagation fail to capture the dynamic relationships between \nagents and their influence on further decision making process. \nTherefore, these models fail to adapt agents to frequent \nenvironment changes. In general, a well-founded trust and \nreputation system that prevents malicious acts that are emerged by \nselfish agents is required for multi-agent systems. We propose a \ntrust mechanism that measures and analyzes the reliability of \nagents cooperating with one another. This mechanism concentrates \non the key attributes of the related agents and their \nrelationships. We also measure and analyze the public reputation \nof agents in large-scale environments utilizing a sound reputation \nmechanism. In this mechanism, we aim at maintaining a public \nreputation assessment in which the public actions of agents are \naccurately under analysis. On top of the theoretical analysis, we \nexperimentally validate our trust and reputation approaches \nthrough different simulations. Our preliminary results show that \nour approach outperforms current frameworks in providing accurate \ncredibility measurements and maintaining accurate trust and \nreputation mechanisms.
Author: Babak Khosravifar
Language: English
Topic: Access Control and Trust
Subfield: Sociology and Political Science
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W2001652