Network Intrusion Detection through Stacked Machine Learning Models on UNSW-NB15 Data Set

M. Karthi, Angela Jeffrin A., Maria Joe Gifta B.

2025

Abstract

Network intrusion detection has become a critical concern in modern cybersecurity due to the increasing frequency and sophistication of cyberattacks. Traditional methods for detecting malicious activities often face challenges when it comes to adapting to novel attack techniques and high-volume data. These methods typically rely on predefined rules or signature-based approaches, which are ineffective against zero-day or unknown attacks. To address these challenges, machine learning (ML) techniques have gained prominence due to their ability to learn from historical data and generalize well to new, unseen attack patterns. However, the performance of individual machine learning models can often be limited due to issues like overfitting, bias, and inability to handle complex feature interactions. This paper introduces an innovative approach that leverages stacked machine learning models for network intrusion detection. Using the UNSW-NB15 dataset, which simulates real-world network traffic with various types of attacks, we combine the strengths of multiple machine learning models Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR) through a technique known as stacked generalization. The base models are trained independently on the dataset, and their individual predictions are used as input features for a meta-model (Logistic Regression), which combines them to make a final, more robust prediction. The stacked generalization technique allows the meta-model to learn the optimal way of combining the base classifiers, thus enhancing the overall performance and making the system more resilient to various types of attacks. By integrating multiple models, this approach capitalizes on the complementary strengths of each individual model, providing a more effective solution for detecting network intrusions. Experimental results show that the stacked model significantly outperforms individual classifiers, achieving an accuracy of 94.1% and demonstrating notable improvements in key performance metrics such as precision, recall, and F1-score. The Random Forest, SVM, and Logistic Regression models each performed well individually, but their combination through stacking resulted in better generalization and improved detection of both known and unknown attacks. This approach not only enhances the detection accuracy but also provides greater robustness against diverse attack vectors present in the dataset. The findings highlight the effectiveness of ensemble learning techniques, particularly stacked generalization, in improving the performance of intrusion detection systems. As cybersecurity continues to evolve, this technique offers a promising direction for building more reliable and adaptive intrusion detection systems capable of addressing the complexities of modern network traffic.

Paper Citation

in Harvard Style

Karthi M., A. A. and B. M. (2025). Network Intrusion Detection through Stacked Machine Learning Models on UNSW-NB15 Data Set. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 538-545. DOI: 10.5220/0013932600004919

in Bibtex Style

@conference{icrdicct`2525,
author={M. Karthi and Angela A. and Maria B.},
title={Network Intrusion Detection through Stacked Machine Learning Models on UNSW-NB15 Data Set},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25},
year={2025},
pages={538-545},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013932600004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25
TI - Network Intrusion Detection through Stacked Machine Learning Models on UNSW-NB15 Data Set
SN - 978-989-758-777-1
AU - Karthi M.
AU - A. A.
AU - B. M.
PY - 2025
SP - 538
EP - 545
DO - 10.5220/0013932600004919
PB - SciTePress