Intrusion detection systems take an important role in securing Internet applications.The exactness of user behavior profiles directly affects the detection performance of intrusion detection systems because profiles are the criterion of anomaly detection.The exactness of profiles would be reduced with the use of traditional profile creation methods due to uncertainty of user behavior patterns in Internet.We proposes a new intrusion detection scheme based on information visualization

and presents a new CCA(Curvilinear Component Analysis)-based visualization algorithm.This algorithm is better than traditional algorithm in the performance of distance mapping

and can provide more exact visual information for security experts.Visual information of user behavior patterns facilitates security experts to select more suitable cluster analysis algorithms to create more exact behavior profiles.