Neural Network-Based Intrusion Detection for Industrial Control Systems in Nigeria’s Petroleum Refining Infrastructure

Industrial control systems governing Nigeria's petroleum refining infrastructure, including the Warri, Port Harcourt, and Kaduna refineries, represent high-value targets for cyber operations whose disruption could generate catastrophic economic and environmental consequences. The application of neural network-based intrusion detection to industrial control system cybersecurity in the Nigerian petroleum sector has not previously been investigated in the published literature. This study develops and evaluates a purpose-designed neural network architecture for intrusion detection in industrial control systems within Nigeria's petroleum refining context, where legacy operational technology systems, network convergence with corporate IT environments, and limited cybersecurity investment create distinct vulnerability profiles. The methodology combines a threat landscape analysis derived from 45 interviews with SCADA engineers, industrial cybersecurity specialists, and Nigerian National Petroleum Company Limited IT security officers with technical model development using simulated industrial control system traffic datasets augmented with Nigeria-specific operational parameters. A bidirectional long short-term memory network with attention mechanisms is developed and trained on the augmented dataset, achieving a detection rate of 97.2 percent for known attack signatures and 89.4 percent for zero-day attack simulations, outperforming baseline models including traditional anomaly detection methods. The study proposes a deployment architecture compatible with the network topologies of Nigeria's refineries and provides a security governance framework for industrial control system cybersecurity aligned with the Nigerian Cybercrime Act 2015. Keywords: intrusion detection, industrial control systems, neural networks, petroleum infrastructure, Nigeria.