Improving Detection of Attacks in Cyber-Physical Systems: Applying Gradient Boosting Based Machine Learning Techniques
Open AccessAs cyber-physical systems (CPS) become increasingly interconnected, ensuring their security against cyber threats is vital. This paper focuses on enhancing the detection of CPS attacks by applying gradient-boosting-based machine learning techniques for intrusion detection systems. Specifically, the performance of gradient boosting models, including XGBoost and LightGBM, is evaluated in classifying different types of cybersecurity attacks using the Edge-IIoTset dataset. The study investigates the influence of various sampling methods on model effectiveness and identifies the key features for accurate intrusion detection. Comparative analyses are conducted with traditional techniques such as K-Nearest Neighbors, Random Forest, and deep learning approaches like Tabnet. The findings consistently demonstrate the superiority of gradient boosting models, exhibiting higher accuracy, and F1-Score in attack classification. Furthermore, it is crucial to underscore the substantial impact of selecting appropriate sampling techniques on model efficacy. It is empirically demonstrated that the use of Random Over Sampling methods tends to enhance model outcomes in a consistent manner. By expanding the scholarly understanding of Cyber-Physical Systems (CPS) defence, this research emphasizes the prospective utilization of gradient-boosting machine learning methodologies to augment intrusion detection measures within CPS. Future research directions include exploring additional ensemble techniques, feature selection methods, real-time implementation, evaluation on diverse datasets, and robustness against adversarial attacks. By addressing these directions, the security and effectiveness of intrusion detection systems in CPS can be further enhanced.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.