SCS Undergraduate Thesis Topics
|Pongsin Poosankam||David Garlan||Authentication and Access Control in Multi-agent Systems|
In a multi-agent system that is dedicated to personal task management, information flows within the system is usually sensitive and should be accessible by only a limited set of people. Some unique properties of the system raise different engineering challenges for the design and implementation of security and access control mechanism: Interpretations of information on different agents have different level of granularity and it can flow through nodes belong to different entities. We propose two design principles for determining when and how access control is enforced:
(1) perform access control as early as possible
(2) by using all information the system can obtained at that time.
In this research, we focus on the RADAR Project as an example of such multi-agent systems. RADAR (Reflective Agents with Distributed Adaptive Reasoning) is a software-based cognitive personal assistant that helps people manage their routine tasks such as answering emails, scheduling meetings, and updating websites. In order to complete the tasks, agents in RADAR communicate with each other to obtain task-related information. In this paper, we describe how we apply the proposed design principles by implementing two levels of information access control policies in RADAR.
The policies are configurable and can be applied efficiently in any multi-agent systems.