Multi-agent Systems (MAS)mustbecapable tocope with continuous requirements changes bywisely making useofemerging components andservices toremain useful intheirenvironment. At thesametime, engineering security into AIIAS confronts uswithnew challenges inrelation tothecomplex andunique characteristics ofagents. Inthis paper, anadaptive and secure MASarchitecture isputforward. Theapproach addresses theneeds ofbothadaptive agent behaviour andtheir security constraints. Central tothemethod is configurable interaction models,in whichthe traditional roleconcept hasbeenextended. Thenew concept incorporates theneedofbothsecurity management asusedbyrole-based access control (RBAC)andagentrole-playing behaviour inagentoriented Software Engineering (AOSE). Themodels capturing the functional and non-functional requirements arecontinuously maintained bybusiness experts atahigh level ofabstraction. Ultimately, agents interpret fromtheup-to-date models their constrained behaviour. Themajorcontribution oftheworkisthat little redevelopment effort willberequired when requirements withregard toagent functions andthe security constraints associated withthefunctions need tobechanged, henceminimising theimpact ofthe requirements changes totheoverall system. A simple expert seeking scenario hasbeenusedtoillustrate the approach.
[1]
A. Poggi,et al.
Multi-User and Security Support for Multi-Agent Systems
,
2001,
WOA.
[2]
Timothy W. Finin,et al.
Modeling conversation policies using permissions and obligations
,
2004,
Autonomous Agents and Multi-Agent Systems.
[3]
Michael N. Huhns,et al.
Making Agents Secure on the Semantic Web
,
2002,
IEEE Internet Comput..
[4]
Leonid Sheremetov,et al.
An authentication protocol for agent platform security manager
,
2003,
EFTA 2003. 2003 IEEE Conference on Emerging Technologies and Factory Automation. Proceedings (Cat. No.03TH8696).