On Physical-Layer Authentication via Triple Pool Convolutional Neural Network
Publication Date
12-1-2020
Document Type
Conference Proceeding
Publication Title
2020 IEEE Globecom Workshops (GC Wkshps)
DOI
10.1109/GCWkshps50303.2020.9367391
Abstract
This paper introduces a novel physical-layer authentication scheme, called Triple Pool Convolutional Neural Network physical-layer authentication (TP-CNN-PHA), aiming to enable a lightweight user authentication mechanism based on physical-layer channel state information (CSI). We first introduce the TP-Net, which is characterized by jointly utilizing maximum pooling, average pooling, and global pooling on a globally connected CNN architecture. To assess its performance, we conduct two sets of experiments, including the one using simulated channel data, and the other one utilizing real experiment data generated from our wireless testbed. The result demonstrates the superiority of the proposed TP-CNN-PHA in terms of authentication accuracy and valid complexity reduction compared with all the considered counterparts, including the threshold-based authentication method.
Funding Number
2018YFB0904900
Funding Sponsor
China Scholarship Council
Keywords
channel state information (CSI), convolutional neural network (CNN), Edge computing, physical-layer authentication
Department
Applied Data Science
Recommended Citation
Yi Chen, Shahriar Real, Hong Wen, Boyang Cheng, Wei Wang, Pin Han Ho, and Shih Yu Chang. "On Physical-Layer Authentication via Triple Pool Convolutional Neural Network" 2020 IEEE Globecom Workshops (GC Wkshps) (2020). https://doi.org/10.1109/GCWkshps50303.2020.9367391
