@Article{cmc.2025.064728,
AUTHOR = {Kuncham Sreenivasa Rao, Rajitha Kotoju, B. Ramana Reddy, Taher Al-Shehari, Nasser A. Alsadhan, Subhav Singh, Shitharth Selvarajan},
TITLE = {Unveiling CyberFortis: A Unified Security Framework for IIoT-SCADA Systems with SiamDQN-AE FusionNet and PopHydra Optimizer},
JOURNAL = {Computers, Materials \& Continua},
VOLUME = {85},
YEAR = {2025},
NUMBER = {1},
PAGES = {1899--1916},
URL = {http://www.techscience.com/cmc/v85n1/63512},
ISSN = {1546-2226},
ABSTRACT = {Protecting Supervisory Control and Data Acquisition-Industrial Internet of Things (SCADA-IIoT) systems against intruders has become essential since industrial control systems now oversee critical infrastructure, and cyber attackers more frequently target these systems. Due to their connection of physical assets with digital networks, SCADA-IIoT systems face substantial risks from multiple attack types, including Distributed Denial of Service (DDoS), spoofing, and more advanced intrusion methods. Previous research in this field faces challenges due to insufficient solutions, as current intrusion detection systems lack the necessary accuracy, scalability, and adaptability needed for IIoT environments. This paper introduces CyberFortis, a novel cybersecurity framework aimed at detecting and preventing cyber threats in SCADA-IIoT systems. CyberFortis presents two key innovations: Firstly, Siamese Double Deep Q-Network with Autoencoders (Siamdqn-AE) FusionNet, which enhances intrusion detection by combining deep Q-Networks with autoencoders for improved attack detection and feature extraction; and secondly, the PopHydra Optimiser, an innovative solution to compute reinforcement learning discount factors for better model performance and convergence. This method combines Siamese deep Q-Networks with autoencoders to create a system that can detect different types of attacks more effectively and adapt to new challenges. CyberFortis is better than current top attack detection systems, showing higher scores in important areas like accuracy, precision, recall, and F1-score, based on data from CICIoT 2023, UNSW-NB 15, and WUSTL-IIoT datasets. Results from the proposed framework show a 97. 5% accuracy rate, indicating its potential as an effective solution for SCADA-IIoT cybersecurity against emerging threats. The research confirms that the proposed security and resilience methods are successful in protecting vital industrial control systems within their operational environments.},
DOI = {10.32604/cmc.2025.064728}
}