FEDERATED META-DEEP LEARNING MODEL FOR REAL-TIME CYBERATTACK DETECTION IN SCADA SYSTEMS | IJCT Volume 13 – Issue 2 | IJCT-V13I2P58
International Journal of Computer Techniques
ISSN 2394-2231
Author
Linda Namikoye Mwibanda, Dr. Lawrence Muriir, Mr. Robert Murungi
Abstract
Supervisory Control and Data Acquisition (SCADA) systems form the backbone of critical infrastructure operations, yet their increasing connectivity has exposed them to sophisticated cyber threats that traditional security mechanisms cannot adequately address. This article presents a novel Federated Meta-Deep Learning (FMDL) model designed to enhance real-time cyberattack detection in SCADA environments. The proposed framework integrates Federated Learning (FL) for privacy-preserving decentralized training, Model-Agnostic Meta-Learning (MAML) for rapid adaptation to novel threats, and hybrid deep learning architectures (CNNs, RNNs, and attention mechanisms) for comprehensive feature extraction. Evaluated using publicly available SCADA datasets (WUSTL-IIoT, SWaT), the FMDL model achieved a detection accuracy of 96.1% ± 1.2, precision of 95.3% ± 1.4, recall of 95.9% ± 1.3, and F1-score of 95.6% ± 1.3, while maintaining a false positive rate of 3.9% ± 0.8. Comparative analysis demonstrated superior performance over traditional centralized models, with a 12.5% reduction in response time (95 ms vs. 110 ms, p < 0.01). User acceptance evaluation using the Technology Acceptance Model (TAM) revealed high scores for perceived usefulness (4.6/5) and behavioral intention to use (4.5/5). The findings establish the FMDL model as a robust, scalable, and privacy-preserving solution for safeguarding critical industrial infrastructure against evolving cyber threats, while acknowledging that real-world validation remains necessary.
Keywords
SCADA security; federated learning; meta-learning; intrusion detection; deep learning; critical infrastructure; cyberattack detection
Conclusion
This study presented a Federated Meta-Deep Learning (FMDL) model for real-time cyberattack detection in SCADA systems. The framework integrates federated learning for privacy-preserving decentralized training, meta-learning for rapid adaptation to novel threats, and hybrid deep learning for comprehensive feature extraction. Evaluated using publicly available SCADA datasets (WUSTL-IIoT, SWaT), the FMDL model achieved high detection accuracy (96.1% ± 1.2), precision (95.3% ± 1.4), recall (95.9% ± 1.3), and F1-score (95.6% ± 1.3), with a false positive rate of 3.9% ± 0.8. Comparative analysis demonstrated superior performance over traditional centralized models, with a 12.5% reduction in response time (95 ms vs. 110 ms, p < 0.01). User acceptance evaluation revealed high scores for perceived usefulness (4.6/5) and behavioral intention to use (4.5/5), indicating readiness for practical adoption.
The FMDL model addresses critical gaps in SCADA cybersecurity by enhancing detection accuracy, ensuring adaptability, minimizing latency, preserving privacy, and gaining user trust. While results are promising, real-world validation remains necessary. As cyber threats to critical infrastructure continue to evolve, frameworks combining decentralized learning with rapid adaptation mechanisms offer promising pathways for strengthening industrial cybersecurity.
References
Ahakonye, A., Nwakanma, C., Lee, J., & Kim, H. (2023). Intrusion detection in SCADA systems using deep learning models: A comparative study. Journal of Industrial Cybersecurity, 10(2), 45-63.
Ahmed, M., & Mahmoud, M. (2022). False alarm reduction in SCADA intrusion detection systems. IEEE Transactions on Industrial Informatics, 18(3), 1842-1851.
Alanazi, M., Mahmood, A., & Chowdhury, M. J. (2023). SCADA vulnerabilities and attacks: A review. Computers & Security, 125, 103028.
Alcaraz, C., & Zeadally, S. (2020). Critical infrastructure protection: Advances and future directions. IEEE Security & Privacy, 18(5), 66-74.
Almuhammadi, S., & Alsaleh, M. (2017). A survey on SCADA systems security: Challenges and solutions. Computers & Security, 70, 436-454.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319-340.
Diaba, S. Y., Nleya, B. D., & Mkhize, T. R. (2023). Neural networks for SCADA security: Analyzing attack patterns and detection efficiency. Neural Networks, 165, 321-332.
Finn, C., Abbeel, P., & Levine, S. (2017). Model-Agnostic Meta-Learning for fast adaptation of deep networks. Proceedings of the 34th International Conference on Machine Learning, PMLR 70:1126-1135.
Ghosh, S., Ganesan, R., & Martinez, J. (2021). Machine learning models for SCADA cybersecurity: A comparative analysis. Applied Sciences, 11(9), 4241.
Hassan, R., Wang, J., & Lin, J. (2022). A survey on machine learning-based SCADA intrusion detection systems. Journal of Cybersecurity Research, 15(2), 223-245.
Kairouz, P., McMahan, H. B., Avent, B., Bellet, A., Bennis, M., & Bhagoji, A. N. (2021). Advances and open problems in federated learning. Foundations and Trends in Machine Learning, 14(1-2), 1-210.
Kenya Power Annual Report. (2023). Kenya Power and Lighting Company Limited Annual Report and Financial Statements.
Kwon, H., Kang, S., & Lee, J. (2019). Anomaly detection in SCADA systems using deep learning. IEEE Access, 7, 112345-112358.
McMahan, H. B., Moore, E., Ramage, D., Hampson, S., & y Arcas, B. A. (2017). Communication-efficient learning of deep networks from decentralized data. Proceedings of the 20th International Conference on Artificial Intelligence and Statistics.
Nguyen, D. C., Ding, M., Pathirana, P. N., & Seneviratne, A. (2023). Federated learning for smart industry. IEEE Transactions on Industrial Informatics, 19(3), 1482-1496.
Riggs, H., Tufail, S., Parvez, I., Tariq, M., Khan, M. A., & Sarwat, A. I. (2023). Impact, vulnerabilities, and mitigation strategies for cyber-secure critical infrastructure. Sensors, 23(8), 4060.
Sharma, N., Patel, R., & Wang, J. (2023). Lightweight machine learning for cybersecurity in industrial systems. IEEE Transactions on Industrial Informatics, 19(4), 3355-3367.
Sharma, P., & Patel, R. (2024). Federated meta-learning for adaptive SCADA intrusion detection. IEEE Transactions on Industrial Informatics, 20(4), 2985-2996.
Wang, Y., Li, X., & Zhang, J. (2024). Adaptive cybersecurity strategies for industrial control systems. International Journal of Critical Infrastructure Protection, 16, 112-127.
Zhang, X., & Li, W. (2021). Decentralized intrusion detection with federated learning in industrial control systems. IEEE Transactions on Industrial Informatics, 17(7), 455-466.
Zhao, Y., Li, M., Lai, L., & Sahu, A. (2021). Federated meta-learning for anomaly detection in industrial IoT networks. IEEE Internet of Things Journal, 8(6), 4396-4405.
How to Cite This Paper
Linda Namikoye Mwibanda, Dr. Lawrence Muriir, Mr. Robert Murungi (2026). FEDERATED META-DEEP LEARNING MODEL FOR REAL-TIME CYBERATTACK DETECTION IN SCADA SYSTEMS. International Journal of Computer Techniques, 13(2). ISSN: 2394-2231.
Related Posts:
