Developing a Risk-Based Cybersecurity Governance Framework for LNG and Tanker Fleet Operations: An OT-Aware Machine Learning and LCGRI Decision Model

Abstract

Liquefied natural gas (LNG) carriers and tanker fleets rely on interconnected operations technology (OT), industrial control systems (ICS), cargo management platforms, power-management systems, navigation equipment, satellite communications, and remote monitoring systems. This digital integration aids operational efficiency and safety monitoring but also opens cyber-physical risk pathways through which a cyber incident can affect propulsion, ballast operations, cargo pumps, monitoring of tank-level contents, power distribution, navigation reliability and emergency situations. While there are useful maritime cybersecurity studies that offer threat taxonomies, regulatory guidance, and intrusion-detection models that are limited work that relates the technical anomaly evidence to governance-level decisions made by fleet managers, designated persons ashore, ship masters, safety auditors, and security officers. This study aims to create a cyber security governance framework for the LNG and tanker fleet operations, based on the HAI Security Dataset as an analytical basis. The HAI dataset was chosen for the features: Hardware-in-the-Loop (HWIL) industrial control system telemetry, labelled attack intervals, multivariate time-series continuity and cyber-physical process coupling. HAI is not one of the datasets generated by the vessels, but it is methodologically appropriate for this study because of the similarities in the OT principles used in LNG and tanker operations: sensor-actuator interaction, process dependency, automated control, and safety-critical monitoring. The research suggests a Liquefied Natural Gas Cyber Governance Risk Index (LCGRI), which is a weighted governance score that incorporates asset criticality, threat probability, vulnerability severity, anomaly-detection confidence, and operational consequence. The suitability of the datasets, class distributions, relationships among the OT features, separability of the anomalies, performance of supervised and unsupervised models, and risk-prioritisation behaviour were all analysed in Python. The results of the analysis indicate that cargo pump manipulation, navigation compromise, cargo tank sensor spoofing, power-management anomalies, and ballast-control disruption are the most important governance issues, while supervised learning models yield consistently strong anomaly-identification performance with HAI-based OT telemetry, with ensemble methods offering additional interpretability advantages for governance. The proposed framework provides a realistic approach to decision support, converting OT anomaly detection to risk-ranked governance actions for LNG and tanker fleet cybersecurity management.