Intrusion detection system (IDS) is an essential tool to ensure security in cyber space. It is meant for detecting deviations in the normal behaviour of the system. An intrusion detection model is used to correctly classify the incoming data as normal or attack. The main objective is to minimize the false positives and enhance the detection rate and classification accuracy. In this paper, an IDS model has been proposed using optimized SVM. The SVM kernel parameters C and γ are optimized using Modified Fruit Fly optimization Algorithm, (M-FOA)in order to increase the detection rate, lower the false alarm rate and improve classification accuracy. The validity of proposed model has been compared with certain state-of-the–art models using KDD CUP’99 dataset. The result shows better performance than the other SVM classifier models in terms of accuracy, detection rate and false alarm rate.