Submit your papersSubmit Now
For Enquiries: [email protected]
IIARD LogoIIARD
📄 Download Paper

Beyond Black Boxes: Permutation Feature Importance for Explainable Intrusion Detection and Model Optimization

Oraye Godspower, N. D. Nwiabu, D. Matthias, E. O. Bennett

Abstract

Deep learning models for network intrusion detection have demonstrated remarkable accuracy, yet their black-box nature limits trust and adoption in security-critical environments. This paper addresses the interpretability gap by applying permutation feature importance analysis to identify the most influential features in intrusion detection. Using our previously established hard voting ensemble of DNN, LSTM, and CNN models (Godspower et al., 2026), we compute permutation importance scores to quantify each feature's contribution to model performance. Experiments on the UNSW-NB15 dataset reveal that only ten features— particularly label, sbytes, smean, sttl, dmean, proto_udp, service_http, synack, id, and sload—account for the majority of predictive power. Retraining the ensemble on these top ten features yields improved computational efficiency while maintaining accuracy (84.98% vs. 85.01%). Feature importance scores also provide actionable insights for network security analysts, enabling focused monitoring of critical traffic attributes. This work demonstrates that permutation importance not only opens the black box of deep learning intrusion detection but also enables model optimization without sacrificing detection accuracy. The findings establish feature importance analysis as an essential tool for deploying trustworthy and efficient intrusion detection systems in operational environments.

Keywords

Permutation feature importanceexplainable AIintrusion detectiondeep learningmodel interpretabilityfeature selectionUNSW-NB15

References

Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5-32. Davies, I. N., Cookey, I. B., & Godspower, O. (2025). Biometric signal processing for security and privacy: Trends, challenges, and future directions. International Journal of Computer Science and Mathematical Theory, 11(11), 76-90. Davies, I. N., Ene, D., Cookey, I. B., Godspower, O., & Deedam, F. B. (2026). An intelligent machine learning framework for password strength and security system. International Journal of Mathematics and Computer Research, 14(4), 6331-6338. Dongmei, Z., Yaxing, W., & Hongbin, Z. (2022). A situation awareness approach for network security using the fusion model. Mobile Information Systems, 2022, Article 6214738. Godspower, O., & Anireh, V. I. E. (2022). Evolution of supercomputer architecture: A survey. International Journal of Computer Science and Mobile Applications, 10(7), 16-26. Godspower, O., Nwiabu, N., & Anireh, V. I. E. (2020). Frequent itemset mining using two dimensional transaction reduction. Journal of Scientific and Engineering Research, 7(7), 7-14. Godspower, O., Nwiabu, N. D., Matthias, D., & Bennett, E. O. (2026). A hard voting ensemble of DNN, LSTM, and CNN for network intrusion detection on the UNSW-NB15 dataset. IIARD , 12(6), 1- 18. Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press. Guyon, I., & Elisseeff, A. (2003). An introduction to variable and feature selection. Journal of Machine Learning Research, 3, 1157-1182. Hassan, M. M., Gumaei, A., Al-Rakhami, M., & Alghamdi, A. (2018). A deep learning approach for network intrusion detection using recurrent neural networks. IEEE Access, 6, 21954-21961. LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. Moustafa, N., & Slay, J. (2015). The UNSW-NB15 dataset: A comprehensive data repository for network intrusion detection systems. In 2015 Military Communications and Information Systems Conference (pp. 1-6). IEEE. Scarfone, K., & Mell, P. (2007). Guide to intrusion detection and prevention systems . NIST Special Publication 800-94. National Institute of Standards and Technology. Sommer, R., & Paxson, V. (2010). Outside the closed world: On using machine learning for network intrusion detection. In 2010 IEEE Symposium on Security and Privacy (pp. 305-316). IEEE. Wang, W. (2024). Cognitive differences in network security based on mathematical modeling based on ensemble learning algorithm. Journal of Cybersecurity and Cognitive Studies, 9(2), 115-132. Zhang, R. C., Zhang, Y. C., Liu, J., & Fan, Y. D. (2019). Network security situation prediction method using improved convolution neural network. Computer Engineering and Applications, 55(6), 86-93. IJCSMT Authors Oraye Godspower received his Bachelor of Science (B.Sc.) degree in Mathematics and Computer Science from Rivers State University of Science and Technology , Port Harcourt, Nigeria, in 2013, and his Master of Science (M.Sc.) degree in Computer Science from Rivers State University , Port Harcourt, Nigeria, in 2021. He completed his Doctor of Philosophy (Ph.D.) in Computer Science at Rivers State University in 2024. He began his academic career as a Graduate Assistant in the Department of Computer Science, Rivers State University, in 2019 and currently serves as a Lecturer II. He is a member of the Computer Professionals of Nigeria . His research interests include network security, intrusion detection systems, deep learning, ensemble methods, and frequent itemset mining. He is the corresponding author and can be contacted at [email protected]. Professor N. D. Nwiabu is a Professor of Software Engineering in the Department of Computer Science, Rivers State University, Port Harcourt, Nigeria. His research focuses on artificial intelligence, cognitive engineering, and their applications to network security and decision support systems. Professor D. Matthias pursued his B.Sc. in Mathematics and Computer Science at the University of Port Harcourt, M.Tech. from the Federal University of Technology Owerri, and Ph.D. from Rivers State University of Science and Technology. He is currently a Professor of Computer Science at Rivers State University. He is a member of CPN (since 2010) and a life member of NCS (since 2012). His research focuses on scientific computing, database systems, and software development. Associate Professor E. O. Bennett graduated with a B.Sc. in Computer Science from Rivers State University (1998), and M.Sc. (2008) and Ph.D. (2014) from the University of Port Harcourt. He is an Associate Professor in the Department of Computer Science, Rivers State University. He is a member of CPN and has published over 50 research papers. His research focuses on algorithms, parallel computing, distributed computing, and intelligent computing.

More Articles from INTERNATIONAL JOURNAL OF COMPUTER SCIENCE AND MATHEMATICAL THEORY

Enhanced Machine Learning-Based Intrusion Detection System for Zero-Day Attacks

Author: UE-Bee, Kakuubari Promise, Sako, D.J.S, Nwiabu, N. D

Virtual Environments of Make Virtual Reality Applications Useful

Author: Ghaliya AlFarsi

Convolutional Neural Network-Based Detection of Facial Image Forgeries Using Binary Classification

Author: Ahmed Samaila1, Etemi. J. Garba2, and Oluborode Kayode O., Corresponding author

Transforming Mathematics Teacher Education Through Artificial Intelligence and Digital Innovation: Extent of Utilization, Challenges, and Strategies in Colleges of Education, Gombe State

Author: Yunusa, Samson PhD, Etuk Emmanuel Dan, Micah Michael, and Ismail Uzairu Usman

Certificate Authentication and Revocation System for IoT Devices using CoAP and CBOR Encoding in LoRaWAN

Author: Duke, Tamunotonye Juliana, Sako, D.J.S., Bennett, E.O.