Zero-Trust Architecture in Hybrid Multi-Cloud Environments: Performance Trade-offs in Digital Banking

Abstract
Modern digital banking infrastructures are rapidly transitioning from legacy perimeter-based security models to hybrid multi-cloud architectures. While this transition enhances operational elasticity and service delivery, it expands the attack surface, requiring the adoption of Zero-Trust Architecture (ZTA) governed by the principle of “never trust, always verify.” However, enforcing continuous identity verification, dynamic mutual TLS (mTLS) handshakes, micro-segmentation, and real-time Policy Decision Point (PDP) evaluations across distributed on-premise mainframes and public cloud providers introduces substantial computational latency and network overhead. This paper presents an empirical performance trade-off analysis of ZTA implementation within high-throughput digital banking systems. Evaluating a dataset of 500,000 real-time payment transactions across a simulated hybrid multi-cloud testbed, we quantify the impact of granular security policies on end-to-end transaction latency, system throughput (Transactions Per Second, TPS), and resource consumption. Empirical results show that an unoptimized, naive ZTA implementation increases mean round-trip transaction latency by 184.2% (from 42 ms to 119.3 ms) and degrades peak throughput by 41.6%. To mitigate these penalties, we propose an Adaptive Edge-Cached Policy Enforcement Framework (AEC-PEF) utilizing localized Policy Enforcement Points (PEPs) and risk-based dynamic token caching. AEC-PEF recovers 72.4% of lost throughput while maintaining an optimal security assurance posture, offering a scalable blueprint for resilient, low-latency financial systems.