The mission of this project is to develop an autonomous, privacy-preserving Edge-AI security node that bridges the gap between sophisticated behavioral analytics and data sovereignty. By leveraging the Hailo-10H NPU for high-throughput skeletal tracking and integrating it with Time-of-Flight (ToF) metric depth data, the system calculates a real-time Behavioral Threat Index (T) entirely on-device. This research focuses on optimizing the late fusion of heterogeneous sensors to detect loitering and spatial anomalies with sub-150ms latency, proving that high-fidelity security monitoring can be achieved without cloud dependency or the sacrifice of individual privacy.