Network Security of Ocean Data Sharing System Integrating Blockchain and Federated Learning

Abstract

This paper combines blockchain and federated learning (FL) technologies to build a data security sharing framework adapted to the marine Internet of Things environment, solve the problems of privacy protection, security assurance and node reliability in the process of marine data sharing, and improve the system's adaptability and data processing efficiency in harsh marine environments. This paper proposes a marine environment model that considers two key factors, the dynamics of ocean currents and the instability of acoustic communication. It quantifies the impact of the marine environment on IoT nodes. The FL framework is applied in the marine IoT environment, and the nodes perform model training locally to predict the changing trend of the marine environment. The FedAvgM (Federated Averaging with Momentum) algorithm is used to improve the training efficiency. Gaussian noise is introduced through differential privacy to prevent sensitive information leakage, and homomorphic encryption technology is used to encrypt model parameters so that they can still be calculated in an encrypted state. The PBFT (Practical Byzantine Fault Tolerance) consensus algorithm is used through blockchain technology (BT) to ensure the high reliability and security of the system. Experimental results show that FedAvgM can achieve an accuracy of 88.2% after 45 iterations. The latency of the PBFT consensus algorithm is 0.1 seconds, and the throughput reaches 2000 TPS (Transactions Per Second). The system that integrates blockchain and FL is superior to the system that uses only a single technology in terms of transmission time, bandwidth utilization, and packet loss rate. The bandwidth utilization and packet loss rate of the system in this paper reach 95% and 1.6% respectively. The system based on the integration of blockchain and FL proposed in this paper can effectively improve the communication efficiency and data security in the marine Internet of Things environment, and can effectively deal with different types of malicious attacks, with strong practicality and application prospects.