Intrusion Detection is one of the most important technique used in the context of security over network. Many tools are available for intrusion detection which use classification for intrusion detection. Accuracy is the most important characteristics to assess the usefulness of the tool. Accuracy of the classifier can be improved by applying feature selection methods. Many of the existing studies illustrate the application of feature selection methods improve accuracy of the model generated by classifier. In order to improve the accuracy of the classification model further in this paper we have proposed hybrid approach for feature selection. Hybrid approach uses the combination of wrapper filter and mutual information measure is used to identify the redundant attributes and remove them before applying classification algorithm.