To protect the information in computer networks, several methods have been designed and implemented.One of these systems is intrusion detection system that uses data mining methods.In this paper, we intend to explore the feasibility of applying feature extraction methods to the misuse detection task.In fact proposed method focuses in data preprocessing.The feature extraction phase is performed by 6 classifiers which we call FE-Classifiers.Assessment of proposed method will be done by KDDCup99 data set.Experimental results exhibit that intrusion detection system with feature extraction method has better performance than that without feature extraction method both in classification rate, detection rate and false alarm rate.