Securing cyber-physical systems with two-level anomaly detection strategy.

Ahmad, Zeeshan; Petrovski, Andrei

Abstract

Cyber-physical system (CPS) represents the integration of digital technologies with physical processes to revolutionize Industry 4.0 by optimizing the industrial processes. However, due to the integration of interconnected devices, the internet, and physical processes, CPS is more susceptible to cyber and physical anomalies. Anomaly detection systems can be implemented to enhance CPS security by actively identifying both physical and cyber irregularities through continuous data monitoring. To this end, this study proposes a two-level detection strategy to secure CPS from all types of anomalies. The first level uses a hybrid Convolutional Neural Network and Long Short-Term Memory to perform the binary classification. Whereas the second level uses a Gradient Boosting Machine to detect the exact type of anomaly. The proposed methodology is evaluated on the physical and network hardware-in-the-loop dataset obtained from a Water Distribution Testbed. The evaluation results demonstrated a high F1-score of 100% and 97.3% on network and physical data respectively, exhibiting its efficiency in accurately predicting anomalies while capturing the most relevant instances to achieve high accuracy.

Citation

AHMAD, Z. and PETROVSKI, A. 2024. Securing cyber-physical systems with two-level anomaly detection strategy. In Proceedings of the 7th IEEE (Institute of Electrical and Electronics Engineers) Industrial cyber-physical systems international conference 2024 (ICPS 2024), 12-15 May 2024, St. Louis, USA. Piscataway: IEEE [online], article number 10639983. Available from: https://doi.org/10.1109/ICPS59941.2024.10639983

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 7th IEEE (Institute of Electrical and Electronics Engineers) Industrial cyber-physical systems international conference 2024 (ICPS 2024)
• Start Date: May 12, 2024
• End Date: May 15, 2024
• Acceptance Date: Feb 25, 2024
• Online Publication Date: Aug 26, 2024
• Publication Date: Dec 31, 2024
• Deposit Date: Aug 1, 2024
• Publicly Available Date: Aug 1, 2024
• Publisher: Institute of Electrical and Electronics Engineers (IEEE)
• Peer Reviewed: Peer Reviewed
• Article Number: 10639983
• Series ISSN: 2769-3899
• DOI: https://doi.org/10.1109/icps59941.2024.10639983
• Keywords: Anomaly detection system; Convolutional neural network; Cyber-physical systems; Gradient boosting machine; Long short-term memory
• Public URL: https://rgu-repository.worktribe.com/output/2423245