SemDPLA：面向6G致密物联网的基于语义通信的分布式物理层认证
SemDPLA: Semantic Communication-based Distributed Physical-Layer Authentication for 6G-enabled Dense IoT
浏览论文内容
中文总结 AI 辅助
针对6G致密物联网中多用户物理层认证指纹可区分性弱和资源受限问题，提出基于语义通信的分布式PLA框架SemDPLA，融合中心与分布式节点语义CSI指纹，结合ArcFace分类和分布式投票机制，在低SNR下实现高精度认证并具鲁棒性。
中文摘要 AI 辅助
随着6G的快速发展，日益密集的设备连接对多用户物理层认证（PLA）提出了更严格的要求。与基于密码学的方法相比，PLA利用无线信道的唯一性实现轻量级认证。然而，现有密集无线场景下的PLA方案往往面临指纹可区分性弱以及计算和通信资源有限的问题。为应对这些挑战，我们提出了一种基于语义通信的分布式PLA（SemDPLA）框架。该框架通过融合来自中心节点和多个分布式节点的语义信息，构建融合中心语义信道状态信息（CSI）指纹。具体而言，我们引入语义通信以降低低信噪比（SNR）的影响以及从分布式节点到中心节点传输过程中的通信资源消耗。此外，我们提出了一种基于ArcFace的分类方法和一种面向语义指纹的分布式投票一致性机制，以提高设备分类精度。仿真结果表明，所提出的SemDPLA方案优于单节点认证、决策融合、原始CSI传输和特征融合基线。它在0 dB和20 dB下分别实现了8.6%和3.4%的等错误率（EER）。它还在0 dB下实现了91.4%的分类精度，在5至20 dB范围内实现了高于95.8%的分类精度。此外，SemDPLA在低SNR环境和异常节点攻击下具有鲁棒性。
英文摘要
With the rapid development of 6G, increasingly dense device connectivity imposes more strict requirements on multi-users Physical-Layer Authentication (PLA). Compared with cryptography-based methods, PLA enables lightweight authentication by using the uniqueness of wireless channels. However, existing PLA schemes in dense wireless scenarios often suffer from weak fingerprint discriminability and limited computation and communication resources. To address these challenges, we propose a Semantic Communication-based Distributed PLA (SemDPLA) framework. The framework constructs fused central semantic Channel State Information (CSI) fingerprints by fusing semantic information from a central node and multiple distributed nodes. Specifically, we introduce semantic communication to reduce the impact of low Signal Noise Ratio (SNR) and the consumption of communication resource during the transmission from distributed nodes to central node. Furthermore, we propose an ArcFace-based classification method and a semantic fingerprint-oriented distributed voting consistency mechanism to enhance device classification accuracy. Simulation results demonstrate that the proposed SemDPLA scheme performs better than single-node authentication, decision fusion, raw-CSI transmission, and feature fusion baselines. It achieves equal error rates (EERs) of 8.6% at 0 dB and 3.4% at 20 dB. It also achieves classification accuracies of 91.4% at 0 dB and above 95.8% from 5 to 20 dB. Moreover, SemDPLA is robust in low SNR environment and under attacks of abnormal nodes.
发表机构
- Beijing University of Posts and Telecommunications（北京邮电大学）
- Xiong’an Aerospace Information Research Institute（雄安航天信息研究院）
- Nanyang Technological University（南洋理工大学）
- University of Surrey（萨里大学）
机构由 AI 辅助整理，请以论文原文为准。
