Performance Evaluation of Intrusion Detection System with SVM

Various machine learning techniques have been proposed to improve the effectiveness of Intrusion Detection Systems (IDS), where IDS is one of the important parts of the network that functions to maintain network security. With its various capabilities and reputation, Support Vector Machine (SVM) is often applied in various classification cases. However, it has constraints with accuracy and computing time when it comes to data that has large dimensions. Utilizing metaheuristic techniques is one possibility that researchers can use to solve this problem. To overcome this, in this study, we will evaluate the performance of SVM for the IDS case by utilizing the metaheuristic Grey Wolf Optimization (GWO) algorithm. From the experiments with three different iteration models, the best results of the combination of SVM and GWO were found in a model with 10 GWO iterations and 10 populations with an accuracy of 97.47% using 15 features out of 48 original features of the UKM-IDS20 dataset.