Physical Layer Spoofing Attack Detection in MmWave Massive MIMO 5G Networks

Abstract

Identity spoofing attacks pose one of the most serious threats to wireless networks, where the attacker can masquerade as legitimate users by modifying its own identity. Channel-based physical-layer security is a promising technology to counter identity spoofing attacks. Although various channel-based security technologies have been proposed, the study of channel-based spoofing attack detection in 5G networks is largely open. This paper introduces a new channel-based spoofing attack detection scheme based on channel virtual (or called beamspace) representation in millimeter wave (mmWave) massive multiple-input and multiple-output (MIMO) 5G networks. The principal components of channel virtual representation (PC-CVR) are extracted as a new channel feature. Compared with traditional channel features, the proposed features can be more sensitive to the location of transmitters and more suitable to mmWave 5G networks. Based on PC-CVR, we offer two detection strategies to achieve the spoofing attack detection tackling static and dynamic radio environments, respectively. For the static radio environment where the channel correlation is stable, Neyman-Pearson (NP) testing-based spoofing attack detection is provided depending on the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${\ell _{2}}$ </tex-math></inline-formula> -norm of PC-CVR. For the dynamic radio environment where the channel correlation is changing, the problem of spoofing attack detection is transformed into a one-class classification problem. To efficiently handle this problem, an online detection framework based on a feedforward neural network with a single hidden layer is presented. Simulation results evaluate and confirm the effectiveness of the proposed detection schemes. For the static radio environment, the detection rate can be improved around 25% with the help of PC-CVR under the NP testing-based detection, and the detection accuracy can reach 99% with the machine learning-based scheme under the dynamic radio environment.

Topics & Concepts