Detecting Malicious Witness Reports in Multi-agent Systems

Abstract

In open Multi-agent Systems, trust plays the central role in facilitating interactions. Most of trust models need to collect the witness reports, and they may suffer due to existence of malicious witness. To detect the malicious witness reports, this paper presents an approach, OSM (Opinion Similarity Measure). Evaluator calculates the opinion similarity state (OSS) between evaluator and witness according to their reports, and evaluates the witness credibility according to the OSS between them. Experiment and analysis show that OSM is more robust than existing approaches.