Protection in a Mobile Agent System

Abstract

Security is perceived as the major obstacle to the widespread acceptance and deployment of mobile agent systems in an open, distributed environment such as the Internet. Two fundamental problems in such systems are, the protection of host resources from agents visiting from untrusted domains, and the protection of agent contents from malicious hosts. This pa.per only addresses the resource protection problem in a mobile agent system implemented using Java objects. This involves the creation of separate protection domains for visiting agents, and granting them controlled access to host resources.We review several approaches for protecting resources, and discuss our design of a proxy-based mechanism for addressing this problem. We use elements of the Java security model and its class-hierarchy based type model in developing this design.