Rollout algorithm for light‐weight physical‐layer authentication in cognitive radio networks. Issue 18 (1st November 2020)

• Record Type: Journal Article
• Authors: Yan, Shengnan; Wang, Xiaoding; Xu, Li
• Is Part Of: IET communications. Volume 14:Issue 18(2020)
• Page Start: 3128
• Page End: 3134
• Publication Date: 2020-11-01
• Subjects: Markov processes -- regression analysis -- cognitive radio -- telecommunication security -- function approximation -- authorisation -- Bayes methods; rollout algorithm -- light‐weight physical‐layer authentication -- cognitive radio networks -- CRNs -- wireless nature -- cognitive nature -- cryptographic authentication -- RoPOMDP -- spoofing attack probability -- nonlinear function approximation -- Markov decision process‐based algorithm -- v‐support vector regression; Telecommunication systems -- Periodicals; Speech processing systems -- Periodicals; 621.38205
• DOI: 10.1049/iet-com.2019.1275
• Languages: English
• ISSNs: 1751-8628

Abstract

Abstract : Cognitive radio networks (CRNs) are vulnerable to spoofing attacks due to their wireless and cognitive nature. Since the traditional cryptographic authentication can hardly prevent such attacks in CRNs, the physical‐layer authentication has been investigated for recent years. To achieve a light‐weight physical‐layer authentication, a rollout partially observable Markov decision process‐based algorithm, named RoPOMDP, is proposed in this study. In general, RoPOMDP formulates the physical‐layer authentication as a zero‐sum game, based on which a hypothesis test upon channel vectors is developed. That allows us to design the gains for both spoofers and receivers based on Bayesian risks for the game, in which the spoofing attack probability is predicted by a non‐linear function approximation utilising v‐support vector regression. Then, a RoPOMDP is employed to estimate the optimal threshold for the test statistic such that spoofing attacks can be detected. The theoretical analysis and simulations indicate that: (i) RoPOMDP improves the spoofing detection accuracy; (ii) as a light‐weight algorithm, the complexity of RoPOMDP is lower than contemporary ones.