A Comprehensive Survey on Federated Learning and Neural Networks for Privacy and Security in Industrial IoT Applications

The development of the Industrial Internet of Things has enhanced the ability to automate tasks. However, the widespread adoption of new devices raises serious questions related to privacy, security, and coverage. Traditional centralized methods of machine learning are inadequate due in part to privacy concerns, communication burden, and attacks. To address this, the notion of Federated Learning has gained traction as a collaborative learning process in which models are trained without raw data ever being shared or exposed. Concurrently, neural networks have been shown to be viable for uses such as intrusion detection, anomaly detection, and malware detection in IIoT systems. This paper explores the relationship between federated learning and neural networks for the protection of privacy and security within IIoT systems. We begin by discussing the Architectural framework of IIoT systems, the potential threats, and the comparative privacy issues present. We then explore the IIoT systems that have proposed federated learning frameworks and outline the pros and cons of these implementations. Additionally, we consider the protection of IIoT security using neural networks and various approaches to its privacy protection.

کلیدواژه ها:

Federated Learning ، Neural Networks ، Industrial Internet of Things (IIoT) ، Privacy-Preserving Machine Learning ، Intrusion Detection Systems