Online Learning Aided Adaptive Multiple Attribute-Based Physical Layer Authentication in Dynamic Environments

• Published in: IEEE transactions on network science and engineering Vol. 8; no. 2; pp. 1106 - 1116
• Main Authors: Yin, Xinyu, Fang, Xiaojie, Zhang, Ning, Yang, Peng, Sha, Xuejun, Qiu, Jinghui
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.04.2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Artificial intelligence; Authentication; Communication system security; Distance learning; False alarms; Heuristic algorithms; Internet of Things; Internet of Things (IoT); Machine learning; multiple attributes; online-learning; Physical layer authentication; Provisioning; Reliability; Search algorithms; security; Signatures; Software radio; Spoofing; Wireless communication
• ISSN: 2327-4697; 2334-329X
• DOI: 10.1109/TNSE.2020.3013232

Exploiting physical (PHY)-layer characteristics for authentication has great potential to provision underlying trust for low-ended Internet of things (IoT) devices with limited computation resources. In this paper, we propose an online learning aided adaptive PHY-layer authentication framework for enhanced authenticity provisioning. Instead of relying on some preset PHY-layer signatures with scenario-sensitive "thresholds," multiple PHY-layer attributes are jointly considered by the proposed scheme to improve the reliability and robustness of PHY-layer authentication. Such a dimension extension on PHY-layer signatures can effectively deteriorate the spoofing capability of malicious attackers. However, it also complicates the predicting and authenticating procedure of the legitimate receiver. Therefore, artificial intelligence aided search algorithms are formulated to facilitate adaptive selection of the Most Effective PHY-layer Attributes (MEA) through learning their historical authenticity performance. To be specific, the attributes which are predicted to be effective in maximizing the authentication capability, in terms of low false alarm rate and low miss detection rate, will be dynamically selected by the authenticator in an autonomous manner. Theoretical analysis shows that the proposed learning algorithm achieves asymptotically diminishing regret. Moreover, extensive experiments are conducted using Universal Software Radio Peripherals (USRPs) in a laboratory environment, which further validates the efficiency of the proposed scheme.