With the increasing modernism in our society, networked computers are playing a pivotal role in dispersion of knowledge, and the protection of critical data in information systems has become a challenge for the research and industrial community. The intrusion detection systems undermine huge amounts of attack data to extrapolate patterns using machine learning techniques. In this paper, a two-stage intrusion detection model has been proposed to employ a blend of diverse attribute selection techniques and machine learning algorithms to provide high performance intrusion detection. The first stage extracts the relevant attributes by applying a hybrid meta-heuristic feature selection algorithm, and in the second stage, supervised machine learning algorithms have been implemented to improve the detection accuracy, execution time, and error rate. NSL-KDD dataset has been used, and the performance of CFS-MHA has been evaluated using different classification strategies. By using 10 attributes and random tree ensemble techniques, CFS-MHA has achieved an accuracy of 81.2% in detection of attacks.