Intrusion Detection Systems (IDS) have become a vital part of computer networks. Specific signatures of formerly identified attacks in a network and characterized traffic datasets are the two most significant parameters, which have been considered by various conventional IDS. Machine learning methods can be applied in IDS since they can learn from attacks' signatures or normal-operations occurring in the network. There is usually a large volume of data in intrusion detection systems in terms of both features and instances. But in this voluminous data, all features do not contribute to traffic thereby increasing the chances of false alarm generation. Therefore, efficiency and veracity of Intrusion Detection Systems can be reduced by selecting only a fair number of features. In this work, an IDS using a recursive feature selection algorithm has been proposed which aids to eliminate various irrelevant features and identify various relevant features of attacks in order to improve attack detection and reduce false alarm generation rate. The proposed IDS has also been analyzed and tested using a revised version of the KDD dataset in Scikit-learn library of Python.