Ensemble Windows Intrusion Detection System using XGBoost and Deep Learning

Abstract

: Intrusion detection systems (IDS) are critical for preserving the Windows environment from an ever-changing collection of cyber threats. This paper introduces XGBNN, a new ensemble model that combines the benefits of deep learning (DL) and machine learning (ML) techniques to identify and mitigate attacks against Windows machines effectively. The various ML methods are trained on the publicly available dataset to classify eight types of attacks in a Windows environment. Additionally, deep neural networks (DNNs) are proposed by optimizing the layers and hyperparameters to achieve the best optimal accuracy. Then, the DNN model and XGboost model are integrated to detect intrusions by utilizing the feature extraction ability of DNN and providing the intermediate features extracted from the last second layer of the DNN to the XGBoost for classification. The Ensemble model XGBNN optimizes features and offers better decisions. The proposed model achieves exceptional accuracy of 100%, as demonstrated by the empirical results, and outperforms the benchmark models. The purpose of this study is to highlight the effectiveness of hybrid architectures in intrusion detection. These architectures offer a more robust, scalable, and effective method to improve the security of the Windows system against more sophisticated attacks.