Abstract: Intrusive behaviors of attackers always imply their objects and intentions, hereby a hierarchical model of intrusion intention recognition was proposed. In order to handle the uncertain message in the circumstance of computer networks, intrusion intention recognition algorithm based on probabilistic inference was presented. On this basis the future plans and goals of the attackers can be predicted in order to take actions in advance. Bayesian networks were constructed to cope with the concurrent intention recognition problems according to the relationship of network security events, goals and intentions. Feasibility and validity of this method were proved from the experiments.

Key words: Intention recognition, Valid intention, Probabilistic inference, Bayesian networks