Abstract: Any intrusion detection system may use both misuse detection and abnormal approach to recognize possible detected attacks. Classification is the problem of intrusion detection. Classification of intrusion detection data is generally divided into two main parts: feature selection and learning algorithms. Various methods have been proposed in connection with feature selection techniques and learning algorithms. The aim in the proposed method is to increase the classification secure and to reach the highest productivity. In present study a hybrid approach is proposed which operates on the combined output of the classifier. The proposed method uses a training set of KDD-Cup99. The proposed method uses three main learning algorithms, SVM, Naive Bayes and J48 decision tree is implemented and evaluated separately. These algorithms are also implemented and evaluated individually as well. The results show the superiority of the proposed method with 97% efficiency using J48 learning algorithm and Adaboost classification by reducing the dimension IG method (feature selection).