A Method Of Hostile Plan Recognition Based On DCSP

Intelligent planning and plan recognition is an important research field in Artificial Intelligence. Plan recognition involves inferring the intentions and plan of an agent from a set of observations and predicts future actions.As a hot research branch of intelligent planning, hostile planning only focus on game strategy domain. The research on hostile planning is still at a preliminary. In hostile planning domain, agent can establish its own plan effectively based on adversarial plan by recognized accurately. Therefore, whether use of plan recognition methods is appropriate decides the strategy of agent directly. There are many limitations in traditional plan recognition approaches, which restrict its application.So we introduce a new plan recognition method based on an observed action set, a dynamic action pending set, a dynamic likely plan set and DCSP, which can recognize adversarial most possible final goal effectively. Our method not only can handle problems under partial observation and partial ordered plans but also can eliminate misleading actions. The system based on our new plan recognition method can be applied in computer network security and intrusion detection.Based on the algorithm proposed above, we have implemented a new plan recognition system, which cannot only recognize the plan and the goal the agent being performed efficiently, but also indicate the actions the agent will take. More over, as there is no ready plan recognition system inland, our system will be significant both in research and practice due to its novelty.