Year: 2003
Type: dissertation
Abstract: In an open multi-agent system (open MAS), agents belonging to the system must use the information they have about each other in order to detect complementary expertise or collect necessary information since the environment is highly dynamic, uncertain, insecure, large, and distributed. The ability to reason about trust will greatly reduce the amount of risk for interacting agents in an open MAS and, in turn, promise to promote trust in the system itself. This research proposes a computational model that combines belief revision and trust reasoning processes (Hybrid Trust Revision combining Direct and Indirect Trust Revision) to develop accurate beliefs about incoming information and information sources within an open MAS. Equipped with the proposed algorithms, an agent is capable of (1) creating accurate and robust beliefs for quality decision-making despite of noise in the incoming information or source reputations, and (2) employing soft security to gradually isolate deceptive or incompetent agents within the system to minimize their impact on decision-making. Although this research primarily focuses on supporting agents in multi-agent systems, it can be applicable to any decision-making system characterized by dynamism, uncertainty, insecurity, scalability, and distributed control.
Author: Joon Woo Kim
Language: English
Topic: Access Control and Trust
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1726103117