Security Enhancement during Agents Communication in GNA Approach

Abstract

In many dynamic environments, where humans and agents coexist, agents cooperate for internal and external resources to achieve their goals. Consequently, the agents are exposed to the risk of being attacked by malicious agents. Such situation, if not mitigated, risks the entire model and threatens its long-Term performances. Due to uncertainties of an agent and its potential behavior, especially in human-Agent collaboration systems, an agent might behave fraudulently to its partners (humans or agents). Therefore, security is a crucial aspect in the initial development of a collaborative human-Agent system. However, many research envisages the security aspects only in the implementation stage. At this stage, a security model is strictly formulated based on the techniques and constraints among agents. Over the past decade, many researchers propose various security models with outstanding features. However, despite the features of the proposed security models, there is still a lack of research effort in security aspects for multi-Agent systems, which is considered as a critical challenge. In this paper, we attempt to enhance the Generic Nodal Abstraction (GNA) approach with a security model applied among agents in their communication. The enhancements include authentication, message encryption and interaction constraints. © 2018 IEEE.