Publication

Abstract : Network intrusion detection is a critical component of cybersecurity systems, aimed at identifying malicious activities within network traffic. With the increasing scale and complexity of modern networks, traditional security mechanisms face limitations in detecting diverse and evolving attack patterns. Learning-based approaches have therefore gained prominence for analyzing network traffic data and distinguishing normal behavior from intrusions. However, achieving consistent detection performance remains challenging due to high-dimensional features and variability in attack characteristics. This paper presents a learning-based intrusion detection framework using the UNSW-NB15 dataset, formulated as a binary classification problem separating normal and attack traffic. The study evaluates classical machine learning models, a deep learning model based on a multilayer perceptron, and hybrid ensemble strategies that combine neural and tree-based learners. Performance is assessed using standard evaluation metrics, including accuracy, precision, recall, and F1-score. Experimental results demonstrate that hybrid ensemble models outperform individual learning approaches by effectively leveraging complementary decision patterns. The findings highlight the effectiveness of combining machine learning and deep learning techniques for robust intrusion detection in contemporary network environments.