Enhancing Countermeasure Techniques in Agent Systems

Multi-agent systems are one of the fastest growing areas in computer science thanks to their synergistic properties and the applicability of their characteristics in both technological and non-technological areas. Virus and other malicious software writers have not ignored this fact and are now either taking advantage of multi-agent systems. In some cases, multi-agent methodologies are used to aid in the creation of malicious software. This paper examine another possibility: a malicious agent is created with the intention of taking advantage of open agent systems willing to execute mobile agents. After these malicious agents have been granted permission to execute, viral code is released. Because the multiagent system is presumably a trusted system, agents that run on it will bypass any current intrusion detection system and/or firewall. This paper proposes an additional method to prevent malicious attacks on agents and multi-agent systems. This enhancement consists of an agent-based sanctioning system that applies penalties to agents and multi-agent systems that behave in an anti-social manner, including malicious behavior.