Abstract

Artificial intelligence is rapidly becoming embedded in software intensive and cyber physical infrastructure, where failures and attacks can propagate across digital and physical domains. Yet reliability engineering, safety engineering, and cybersecurity are still frequently practiced as parallel disciplines with different artifacts, metrics, and governance cadences. This paper proposes URSE, a Unified Reliability and Security Engineering framework that connects AI lifecycle controls, software reliability practices, and cyber physical infrastructure protection into one closed loop system. URSE introduces a shared system model, joint assurance artifacts, and a risk weighted reliability objective that aligns operational SRE metrics with infrastructure security outcomes. The framework is operationalized through decision intelligence loops that continuously translate telemetry into architecture level actions: control selection, test prioritization, automated rollback, and policy enforcement. We present a reference architecture and an evaluation design using a representative smart infrastructure scenario to show how URSE reduces mean time to recovery under combined fault plus attack conditions while improving measurable service availability and safety margin.