Architecture Supporting Security of Agent Systems

Abstract

The thesis proposes a new security architecture supporting anonymity of agent systems. The Module I of the architecture provides functionality of untraceability protocol with the advantage of not imposing restrictions on agents autonomously selecting their route. It forms a core of the anonymity architecture, developed as a stand-alone extension of JADE agent platform. To evaluate its performance a new performance evaluation framework was built and multiple studies were performed. The extension was also successfully applied to an e-Health case study. The security analysis of the architecture proved that it satisfies its requirements. At the same time it exposed that powerful adversaries have at disposal a large variety of traffic analysis attacks which can break security provided by standard untraceability protocols. As a result of this observation Module II: Additional Untraceability Support was proposed. This part of the architecture addresses the complex issue of extended untraceability of agents including traffic analysis and tracing agents through reading data held by them. The module was specified as an abstract architecture, through descriptions of its components. The decision of their implementation and application to particular agent systems must be left to administrators and/or designers of the systems.