Research on a New Lightweight Authentication Technology for Power Communication Networks Based on Reconfigurable Intelligent Surface and Meta-Learning

With the expansion of new power communication networks, massive single-antenna power Internet of Things (IoT) terminals face severe security access challenges. Conventional cryptography-based authentication can impose non-negligible computational, memory, and energy overhead on resource-constrained devices, while multi-antenna-aided Physical-Layer Authentication (PLA) cannot generate distinguishable spatial features to resist Sybil attacks for single-antenna nodes. Moreover, heterogeneous power terminals and complex environments lead to high labeling costs and poor adaptability in deep learning models. This paper proposes a lightweight authentication scheme combining a Reconfigurable Intelligent Surface (RIS), Model-Agnostic Meta-Learning (MAML), and Convolutional Neural Network-Gated Recurrent Unit (CNN-GRU) network. RIS optimizes reflection coefficients under unit-modulus constraints to construct distinguishable spatial channel fingerprints. MAML enables fast adaptation with limited samples through the meta-learning framework, enhancing scalability. The lightweight CNN-GRU efficiently extracts spatial and temporal features for edge-side real-time inference. Simulation results verify its superior authentication performance and robustness against time-varying channels and the evaluated Sybil-related attack scenarios, including power mimicry, proximity impersonation, and dynamic-environment interference; other physical-layer threats such as pilot spoofing, replay, and intentional jamming have not yet been comprehensively evaluated. This work contributes to secure, resilient, and sustainable smart grid operations within the evaluated scope.