Privacy-Preserving Multi-Modal Threat Detection Framework for Educational Facilities: A Conceptual Design

Abstract

School safety systems must detect threats rapidly while protecting student privacy and minimizing false alarms. This paper presents a conceptual framework combining edge-based open-vocabulary object detection with millimeter-wave radar sensing. Our three-tier confidence system matches response intensity to detection certainty, incorporating teacher verification at intermediate levels. The framework processes all data locally and includes automatic face anonymization, minimal retention policies, and no identity tracking. We provide practical camera placement methodology using pixels-per-meter calculations and detailed implementation guidelines. While this work presents a design proposal rather than experimental validation, it offers a comprehensive blueprint balancing security effectiveness with ethical requirements. The framework is designed to be reproducible and adaptable to diverse educational environments.