TEJAS Journal of Technologies and Humanitarian Science

Oum Pratap Singh
Research Scholar, Department of Computer Science, Dr. Bhimrao Ambedkar University, Agra

Author

Dr. Sandeep Kumar Jain
Professor, Department of Computer Science, Dr. Bhimrao Ambedkar University, Agra

Author

📜 License:

This work is licensed under a Creative Commons Attribution 4.0 International License

• Share — Copy and Redistribute the material
• Adapt — Remix, Transform, and build upon the material
• The licensor cannot revoke these freedoms as long as you follow the license terms.

Abstract:

Cyber-Physical Systems (CPS) are increasingly deployed in critical infrastructure, making CPS security a priority for both industry and academia. One of the key defenses is the Intrusion Detection System (IDS), which leverages AI and machine learning to identify malicious activities. However, the high dimensionality of modern network datasets often leads to redundancy, increased computational cost, and reduced efficiency in IDS performance. This study presents a comparative analysis of feature selection techniques applied to the UNSW-NB15 dataset using a Random Forest classifier. The methods evaluated include Chi-Square, ANOVA F-test, Random Forest importance, and an all-features baseline. Experiments were conducted on a stratified subset with 2-fold cross-validation to ensure robustness. Results show that Random Forest importance with the top-30 features achieved the best balance of accuracy, macro-F1, and recall, outperforming statistical filter-based approaches while reducing computational overhead. The findings confirm that effective feature selection can enhance IDS performance without compromising detection accuracy, making AI-driven IDS models more suitable for real-time CPS security applications. This research highlights the role of machine learning in building efficient, reliable, and scalable IDS frameworks for securing CPS, and suggests future exploration of hybrid feature selection methods to improve generalization and interpretability.

References

[1] M. A. Ferrag et al., “A systematic review of data mining techniques for intrusion detection systems,” J. Inf. Secur. Appl., vol. 50, 102419, 2020. [Online].
[2] I. Ahmed et al., “A survey of cyber-physical systems intrusion detection datasets,” Computers, vol. 9, no. 2, p. 22, 2020.
[3] N. Shone et al., “A deep learning approach to network intrusion detection,” IEEE Trans. Emerg. Topics Comput. Intell., vol. 2, no. 1, pp. 41–50, 2018.
[4] Z. Chiba et al., “A novel architecture combined with feature selection for effective intrusion detection in cloud computing,” Secur. Commun. Netw., vol. 2019, 8324012, 2019.
[5] G. Kocher and G. Kumar, “Analysis of ML algorithms with feature selection for IDS using UNSW-NB15,” IJNSA, vol. 13, no. 1, pp. 13–
27, 2021.S. M. Kasongo and Y. Sun, “Performance analysis of IDS using feature selection on UNSW-NB15 dataset,” J. Big Data, vol. 7, no. 105, 2020.
[6] Y. Yin et al., “IGRF-RFE: A hybrid FS method for MLP-based IDS on UNSW-NB15 dataset,” arXiv preprint arXiv:2203.16365, 2022.
[7] A. M. Alsaffar et al., “Shielding networks: Enhancing IDS with Boruta FS for CPS security,” J. Big Data, vol. 11, no. 29, 2024.
[8] Y. Zhang et al., “Network intrusion detection: Evaluating FS and ML classifiers on UNSW- NB15 and CICIDS2017,” Sensors, vol. 21, no. 21, 7320, 2021.
[9] N. Shone et al., “A deep learning approach to network intrusion detection,” arXiv preprint arXiv:1804.09534, 2018.
[10] T. A. Tang et al., “Deep learning approach for network intrusion detection in software defined networking,” 2016 Int. Conf. WINCOM, 2016.
[11] Z. Chiba et al., “A novel architecture combined with feature selection for effective intrusion detection in cloud computing,” Secur. Commun. Netw., 2019.
[12] I. Aljawarneh, M. Aldwairi, and M. B. Yassein, “Anomaly-based intrusion detection system through feature selection analysis and hybrid efficient model,” J. Comput. Sci., vol. 25, pp. 152–160, 2018.
[13] A. Javaid, Q. Niyaz, W. Sun, and M. Alam, “A deep learning approach for network intrusion detection system,” Proc. 9th EAI BIONETICS Conf., 2016.
[14] I. Sharafaldin, A. H. Lashkari, and A. A. Ghorbani, “Toward generating a new intrusion detection dataset and intrusion traffic characterization,” ICISSP, 2018.