Application of the AHP method in modeling the trust and reputation of software agents

Given the unique characteristics of cyberspace and, in particular, the number of inherent security threats, communication between software agents becomes a highly complex issue and a major challenge that, on the one hand, needs to be continuously monitored and, on the other, awaits new solutions addressing its vulnerabilities. An approach that has recently come into view mimics mechanisms typical of social systems and is based on trust and reputation that assist agents in deciding which other agents to interact with. The paper offers an enhancement to existing trust and reputation models, involving the application of the AHP method that is widely used for decision support in social systems, notably for risks analysis. To this end, it is proposed to expand the underlying conceptual basis by including such notions as self-trust and social trust, and to apply these to software agents. The discussion is concluded with an account of an experiment aimed at testing the effectiveness of the proposed solution.