Federated Learning-based Anomaly Detection System for Securing Nigerian Oil Pipeline Sensor Networks

Abstract

Oil pipeline infrastructure plays a critical role in Nigeria’s economy by supporting the transportation of crude oil and petroleum products across the country. However, the pipeline network is frequently threatened by vandalism, oil theft, and operational failures, leading to significant environmental damage and economic losses. Wireless Sensor Networks (WSNs) have recently been adopted for real-time monitoring of pipeline parameters such as pressure, temperature, and flow rate. Despite their effectiveness, WSN-based monitoring systems remain vulnerable to cyber-attacks and data manipulation due to their wireless communication architecture. This study proposes a federated learning-based anomaly detection system to enhance the security and reliability of WSNs used in Nigerian oil pipeline monitoring. The proposed framework enables distributed machine learning where sensor nodes locally train models and share model updates rather than raw data, thereby preserving data privacy and reducing communication overhead. Three machine learning algorithms Support Vector Machine (SVM), Logistic Regression (LR), and K-Nearest Neighbors (KNN) were integrated within the federated learning framework to detect anomalies in pipeline sensor data. Simulation experiments were conducted to evaluate the system using performance metrics including accuracy, precision, recall, and F1-score. The results demonstrate that the proposed federated learning model improves anomaly-detection performance while maintaining data privacy and scalability, compared with centralised monitoring approaches.