PhD in Computer Science - Dealing with Miscreant Agents in Orono United States | University of Maine

Agent-based systems. An agent is an autonomous entity that observes its environment, evaluates possible actions, and executes decisions. Multiagent systems (MAS) consist of software or combined software/hardware architectures centered around these intelligent agents. The study of MAS has grown significantly in importance for managing robot teams and coordinating interacting software agents.

When agents act against the system. Traditionally, MAS research assumes agents operate cooperatively or at least neutrally. However, when self-interested agents are introduced, there's a genuine possibility an agent may prioritize its own objectives over collective or group interests. Such agents might be labeled as rogue agents, even without explicit malicious intent toward the system.

Detection and prevention. The CA-SLAM (Context-Aware Society-Level Analysis of Motives) initiative examines systems vulnerable to such problematic behavior. Key research focuses include identifying system vulnerabilities, developing detection methods for rogue agents, and creating system designs that inherently prevent such issues. CA-SLAM originated as a doctoral project at MaineSAIL, with the researcher now continuing this work at UMA.