A Federated Learning-Based Hierarchical Intrusion Detection System for Underwater IoT Wireless Sensor Networks

Underwater Internet of Things Wireless Sensor Networks (UIoTWSN) are an important component in the process of marine monitoring and gathering of environmental data. Due to the dynamic nature of underwater environments and limits of typical intrusion detection system (IDS), these networks are extremely susceptible to incursions. Both of these factors make them extremely vulnerable. The approaches that are currently in use include problem areas such as computational inefficiencies, high false positive rest and lack of adaptability to changing attack patterns. A Federated Learning based Hierarchical IDS for utility IOT is suggested to overcome these issues. By utilizing federated learning with multi-tiered detection framework, this system is able to protect the privacy of data, reduce the amount of computing overhead and effectively respond to emerging threats in real time. The hybrid method employs Convolutional Neural Network (CNN) for the extraction of spatial features and Bi-LSTM networks for the analysis of temporal features. Moreover, reinforcement learning is utilized to dynamically modify network conditions and attack patterns. To maximize the utilization of resource in the underwater sea, an optimization method is utilized. This method is energy constrained, lightweight and energy efficient. In addition, the stability of the method is further enhanced by capability of blockchain technology to allow secure collaboration between nodes. The findings of the experiment shows that the proposed FL-HIDS is a reliable option for ensuring the safety of UIoTWSN since it achieves higher level of accuracy, increased energy efficiency and less false positives in compared to methods that are currently in development. This paper presents an innovative intrusion detection system which is suitable for issues in Underwater IoT