Self-Protection Model Based on Intelligent Agent

In this paper, we propose a self-protection model based on distributed intelligent agents. The major feather of this model is that autonomic agents can manage themselves given high-level objectives from administers. System can detect attacks and compromise, and recovering from security incidents without human intervention. We describe the model and the existing prototype, as well as some design and implementation experiences. Meanwhile several topics including structure of agent, relationship among agents and the establishment of a trustworthy system are discussed. We are implementing this model on the Eclipse Platform using Java. According to the experiments, this model provides significant advantages in scalability, flexibility, fault tolerance, and resistance to compromise.