An Integrated Federated Learning Framework with Optimization for Industrial IoT Intrusion Detection

Abstract

Because of the rapidly growing number of connected devices, the Industrial Internet of Things (IIoT) has presented significant security challenges. IoT networks must be secure and efficient in order to remain sustainable. We present an integration of federated learning and optimization techniques to detect intrusions in IIoT. By combining CNNs with the proposed framework, model accuracy can be enhanced while communication costs are minimized. Data security and user privacy are ensured through the implementation of differential privacy mechanisms. In addition to high rates of accuracy, precision, and recall, the framework has an efficient training management process. With its robustness and scalability, the proposed method offers a better solution for IIoT security concerns than traditional centralized machine learning methods.