An Optimized Machine Learning Framework for Attack Detection in IoV for Marine Applications Using Federated Learning

Abstract

The Internet of Vehicles has transformed marine applications through better connected technology and automated systems, but these systems suffer from substantial vulnerability to online attacks that undermine security and reduce performance. Traditional ways of detecting attacks from centralized systems create performance delays with high computations and privacy threats which are severe issues in marine areas with poor infrastructure. Our research creates an enhanced machine learning (ML) approach through Federated Learning (FL) for secure attack discovery in IoV marine systems. The framework enables decentralized model training across multiple nodes without sharing raw data, preserving privacy, and minimizing bandwidth usage. Advanced feature engineering and optimization techniques are employed to improve the detection accuracy and computational efficiency of the FL-based model. Comprehensive experiments are conducted on CICIDS2017datasets to evaluate the framework’s performance in detecting diverse cyber threats, including spoofing, DoS attacks, and so on. Results demonstrate significant improvements in A_accuracy, P_preecision, R_recall, and F1_score of 98.8% compared to conventional centralized methods. This study highlights the potential of FL-driven approaches to enhance cybersecurity in IoV, offering a scalable and privacy-preserving solution for secure marine operations.