RF Fingerprint Recognition in Resource-Constrained Edge Computing

RF fingerprinting is a method to identify different wireless devices based on hardware differences in the communication devices, aiming to assist in solving the problem of secure access to wireless networks. Traditional RF fingerprinting relies on feature extraction and the training and prediction of data under a single node. However, in real-world application scenarios, it is impractical to send all data to a centralized location, and these approaches do not take into account the limited resources such as bandwidth and storage in edge networks. In this paper, we use distributed machine learning to solve the problem of learning models from data from multiple edge nodes. And the high communication cost in distributed machine learning is optimized for the problem of resource constraint of edge devices. Simulation results show that the method can not only cope with the training problem of a large amount of data but also reduce the communication cost on the edge devices, which can reduce the training time by about 9%.