Radio-Frequency Fingerprinting for Distributed IoT Networks: Authentication and QoS Optimization

With the advance of the Internet of Things (IoT), limited wireless resources cannot meet the requirements of all users. User authentication can identify the users who are not authenticated and only allocate the resources to authenticated users. However, traditional authentication methods require complicated computations that restrict the wide deployment of low-cost devices in IoT applications. This article proposes a federal learning-based authentication method with radio-frequency fingerprinting recognition under a distributed architecture, which is novel and different from the previous works since these works primarily focus on the network with a centralized node. We propose a distributed optimization problem in a hybrid local and edge computing scenario to minimize the network delay. By transforming the original problem into its Lagrange function and calculating its derivatives, we propose a method of solving the optimization problem and prove that the proposed method could reduce the unnecessary consumption of wireless resources and effectively improve the entire network's performance. The analytical results also indicate that the proposed method could outperform the benchmark authentication methods in reducing network delay.