Research Article
Federated Learning Based Distributed Algorithms for RF Fingerprinting Extraction and Identification of IoT Devices
@INPROCEEDINGS{10.1007/978-3-030-90196-7_1, author={Weiwei Wu and Su Hu and Yuan Gao and Jiang Cao}, title={Federated Learning Based Distributed Algorithms for RF Fingerprinting Extraction and Identification of IoT Devices}, proceedings={Artificial Intelligence for Communications and Networks. Third EAI International Conference, AICON 2021, Xining, China, October 23--24, 2021, Proceedings, Part I}, proceedings_a={AICON}, year={2021}, month={11}, keywords={RF fingerprinting Federated learning IoT devices}, doi={10.1007/978-3-030-90196-7_1} }- Weiwei Wu
Su Hu
Yuan Gao
Jiang Cao
Year: 2021
Federated Learning Based Distributed Algorithms for RF Fingerprinting Extraction and Identification of IoT Devices
AICON
Springer
DOI: 10.1007/978-3-030-90196-7_1
Abstract
With the development of Internet of things (IoT), exponential data growth and diversified functions and services have dramatically enhanced the importance of user authentication for data access. As a solution to the problem of user authentication, we study the deep-learning based methods for the radio frequency (RF) fingerprinting recognition of mobile devices in this paper. In consideration of the distributed storage of RF signals in practice, instead of using the deep learning algorithms for centralized data training, we employ the federated learning algorithms for distributed RF fingerprinting recognition, where the data of RF signals are distributed in multiple mobile devices for storage and recognition. To reduce the impact of uneven data distribution among mobile devices on the performance of federated learning algorithms, we propose the dynamic sample selection based federated learning algorithms to train the data. In comparison with the traditional federated learning algorithms, our proposed algorithm can improve the system accuracy as well as reduce the computation time.
