Zero-Trust Architecture for Shared AI Infrastructure: Enforcing Security at the Storage-Network Edge

Abstract

The push towards shared Artificial Intelligence (AI) infrastructure on the multi-tenant and distributed scale poses complicated security predicaments, especially at the data storage and edge networking crossroads. The old perimeter-based security privileges cannot serve these dynamic systems anymore because they need to protect sensitive data and computation processes. This paper proposes an innovative Zero-Trust Architecture (ZTA) implementation that utilizes strong security restrictions at the edge of the storage network to safeguard shared AI infrastructure. The suggested approach combines identity-based authentication, real-time verification, microsegmentation, and dynamic policy enforcement to reduce insider attacks, unauthorized access, and data leakage. The research shows that the framework provides significant gains in data confidentiality, access control granularity, and general resistance to advanced persistent threats when deployed in a containerized AI testbed, as compared to traditional security settings. The results support the use of ZTA as a paradigm for building the future of edge-enabled AI systems, particularly in industries where trust is paramount, such as healthcare, finance, and defense.