An Investigation on Intrusion Detection System Using Machine Learning

With prevalent technologies like Internet of Things, Cloud Computing and Social Networking, large amounts of network traffic and data are generated. Hence, there is a need for Intrusion Detection Systems that monitors the network and analyzes the incoming traffic dynamically. In this paper, NSL-KDD is used to evaluate the machine learning algorithms for intrusion detection. However, not all features improve performance in a large datasets. Therefore, reducing and selecting a particular set of features improve the speed and accuracy. So, features are selected using Recursive Feature Elimination (RFE). We have conducted a rigorous experiment on Intrusion Detection System (IDS) that uses machine learning algorithms, namely, Random Forest and Support Vector Machine (SVM). We have demonstrated the comparison between the model's performance before and after feature selection of both Random Forest and SVM. We have also presented the confusion matrices.