Dynamic Authentication and Granularized Authorization With a Cross-Domain Zero Trust Architecture in Large-Scale IoT Networks

With the increasing number of connected devices and complex networks involved, current domain-specific authentication mechanisms become inadequate for the dynamic access demands of large-scale Internet of Things (IoT) systems. Cross-domain authentication and authorization could enhance device interoperability across different security domains. However, they face new challenges including increased authentication latency, non-transferable trust, and excessive authorization computational overhead. To address these issues, in this paper, we propose Dynamic Authentication and Granularized Authorization (DAGA), a cross-domain Zero Trust Architecture (ZTA)-based framework that enables efficient, adaptive, and scalable authentication and authorization for large-scale IoT systems. Rather than relying on full authentication for every request, DAGA introduces a cross-domain pre-authentication phase that estimates device trustworthiness in advance. Specifically, DAGA leverages Federated Learning (FL) and Long Short-Term Memory (LSTM)-based dynamic trust management to minimize unnecessary trust score computations, thereby accelerating cross-domain pre-authentication. At the same time, DAGA employs multi-level and cached fast authorization to achieve adaptive granularized access control, enhancing the accuracy and efficiency. Experimental results demonstrate that DAGA achieves notable reductions in communication and computation overhead compared to ZTA, blockchain-based, and lightweight mechanisms, significantly reducing authentication latency. Despite its efficiency gains, DAGA achieves a superior accuracy–latency trade-off over other baseline approaches, effectively balancing security and efficiency through adaptive optimization.