Stacked Ensemble-IDS Using NSL-KDD Dataset

Abstract

The intrusion detection system is a traffic monitoring unit that protects our network from numerous threats. It works as a monitoring unit with the ability to detect attacks in a real-time environment. Various techniques have been employed to make the IDS machine work with accuracy. To attain good accuracy, machine learning and deep learning are being utilized to train and evaluate the IDS machines. To prepare for the IDS's training and testing, before deploying them in real-time situations, a collection of real-world internet traffic records is stored with their traffic input features, this traffic records can be utilized to test the IDS machine against several attacks before they are employed into the real world. One such dataset is the University of New Brunswick's NSL-KDD dataset used to train and test the machine learning or deep learning models. In this paper, we will train and evaluate our model using the proposed stacked ensembled machine learning model, and with the NSL-KDD dataset, compare it based on various evaluation metrics against standard machine learning methods and some earlier proposed research.