Randomized CNN based deep learning technique for the cyber-attacks detection in SCADA industrial control systems

Mubarak, Sinil and Habaebi, Mohamed Hadi and Islam, Md. Rafiqul and Jaleel, Nubila and Siddique, Mohammed Tahir (2025) Randomized CNN based deep learning technique for the cyber-attacks detection in SCADA industrial control systems. Measurement, 254 (NA). pp. 1-20. ISSN 0263-2241

	PDF Restricted to Registered users only Download (12MB)
	PDF Restricted to Registered users only Download (187kB)

Abstract

The increasing digitization and connectivity of Industrial Control Systems (ICS) have exposed them to highly sophisticated cyber threats to a great extent. Traditional security mechanisms like rule-based and signature-based intrusion detection systems fail to detect new and emerging attacks. In such limitations, the current research is on the Randomized Convolutional Neural Network (R-CNN) model for cyber-attack detection in ICS networks. The proposed model leverages convolutional layers to improve feature extraction and reduce the likelihood of overfitting. Additionally, advanced data preprocessing, augmentation, and hyperparameter optimization methods to maximize classification performance. With the increasing frequency of cybersecurity attacks in the Industrial Internet of Things (IIoT), addressing such challenges are hindered by outdated public datasets and scarcity of testbeds, to design appropriate solutions to detect and prevent cyberattacks. To overcome this limitation, we have developed an in-house cyber testbed based on standard industrial operations, simulated with hacking scenarios. The model is trained and tuned with ICS cyber-attack benchmark datasets on the various attacks. Randomized layers provide controlled variability at training time, enhancing the model to recognize known and unknown attacks. Experimental results indicate that the proposed R-CNN method performs better than the baseline machine learning and conventional CNN methods with a detection accuracy of 98.7%. The model also achieves 98.2% precision, 97.9% recall, and an F1-score of 98.0%, making the threat detection robust and the false positive rate imperceptible. The computational overhead of the architecture also enables real-time deployment within an industrial setup. The improved Randomized CNN-based IDS in deep learning-based cyber security applications facilitates real-time cyber threat detection feasibility within ICS. It enhances the model for zero-day attack detection and integration with adaptive security paradigms for predictive threat blocking.

Item Type:	Article (Journal)
Uncontrolled Keywords:	Cyber-attacks; CNN; Random CNN; RNN; LSTM; SCADA
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800 Electronics. Computer engineering. Computer hardware. Photoelectronic devices
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Engineering > Department of Electrical and Computer Engineering
Depositing User:	Dr. Mohamed Hadi Habaebi
Date Deposited:	03 Jun 2025 09:38
Last Modified:	16 Jun 2025 14:32
URI:	http://irep.iium.edu.my/id/eprint/121301

Actions (login required)

View Item