Title: Network intrusion detection enhancement system through machine learning algorithms

Authors: Nour Kadhim Hadi

Addresses: Department of Computer Networks and Cybersecurity Engineering, College of Computer Engineering, University of Technology, Baghdad, 10001, Iraq

Abstract: In today's digitally driven landscape, network security is critical as decision-making relies heavily on reliable, high-quality data. This study utilises the network security laboratory-knowledge discovery data mining (NSL-KDD) dataset to improve traffic classification and data quality evaluation through state-of-the-art techniques. By employing advanced methods such as feature engineering, ensemble learning, and machine learning (ML) algorithms, the research enhances classification precision and accuracy. The primary goal is to develop an optimised classifier capable of detecting anomalous network traffic indicative of security intrusions. Using a preprocessed dataset, the study identifies significant features and optimises hyperparameters to refine performance. A decision tree classifier was evaluated using unseen data, integrating stages of data preparation, anomaly detection, training, and visualisation. The resulting model demonstrated high effectiveness for intrusion detection systems, achieving an accuracy of 95.36%, alongside 97% precision, a 95% F1-score, and 95% recall. This comprehensive approach ensures robust detection of potential network threats.

Keywords: machine learning; intrusion detection; decision tree; NSL-KDD; feature importance; performance metrics.

DOI: 10.1504/IJCNDS.2026.154606

International Journal of Communication Networks and Distributed Systems, 2026 Vol.32 No.4, pp.378 - 397

Received: 15 Jul 2024
Accepted: 09 Sep 2024
Published online: 07 Jul 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article