Skip to Main Content
Multi-agent systems (MAS) must be capable to cope with continuous requirements changes by wisely making use of emerging components and services to remain useful in their environment. At the same time, engineering security into MAS confronts us with new challenges in relation to the complex and unique characteristics of agents. In this paper, an adaptive and secure MAS architecture is put forward. The approach addresses the needs of both adaptive agent behaviour and their security constraints. Central to the method is configurable interaction models, in which the traditional role concept has been extended. The new concept incorporates the need of both security management as used by role-based access control (RBAC) and agent role-playing behaviour in agent-oriented software engineering (AOSE). The models capturing the functional and non-functional requirements are continuously maintained by business experts at a high level of abstraction. Ultimately, agents interpret from the up-to-date models their constrained behaviour. The major contribution of the work is that little redevelopment effort will be required when requirements with regard to agent functions and the security constraints associated with the functions need to be changed, hence minimising the impact of the requirements changes to the overall system. A simple expert seeking scenario has been used to illustrate the approach.