Never trust, always verify: a zero trust security framework for internet of robotic things ecosystems

As the Internet of Robotic Things (IoRT) expands across diverse domains, its increasing complexity and distributed nature demand advanced security approaches beyond traditional perimeter-based defenses. Zero Trust Architecture (ZTA) provides a robust cybersecurity framework by enforcing continuous trust evaluation, ensuring that no user, device, or application is inherently trusted. This paper presents a comprehensive IoRT threat model and a novel Zero Trust Authorization Framework for securing dynamic IoRT environments. Aligned with Department of Defense (DoD) Zero Trust guidelines, the framework incorporates weighted trust scoring algorithms to assess risks associated with network behavior, sensor data integrity, and critical system operations. A case study uses a RoboDK-based smart home simulation, where autonomous robots and interconnected devices perform routine tasks under threat scenarios, including unauthorized access, operational faults, and malicious activities. To further evaluate robustness, experiments were conducted using the TON_IoT dataset, which captures diverse, realistic, and evolving IoT attack behaviors. Large Language Models (LLMs) interpret trust scores, evaluate decision thresholds, and dynamically enforce access control based on contextual risk. Overall, experimental results demonstrate that the proposed approach significantly achieves 98.77% accuracy and 99.11% attack recall on the TON_IoT dataset, while improving anomaly detection performance (R2=97.4%), reducing mean absolute error to 15.5%, and strengthening resilience against emerging IoRT-specific threats.