AI-based system for threat detection in academic environments

Abstract

This article investigates the use of artificial intelligence (AI) to enhance security in educational environments through real-time automated threat detection, focusing on the identification of firearms. The proposed system leverages the YOLO architecture, achieving an average precision of 0.994 in controlled tests, demonstrating potential for rapid and effective interventions. Despite challenges such as limited dataset representativeness, the need for robust hardware, and computational resources, the model offers a promising solution for preventing violent incidents in schools. Additionally, ethical and legal aspects, such as compliance with data protection regulations (e.g., Brazil's General Data Protection Law - LGPD) and bias reduction, were integrated into the system's design, reinforcing its practical viability and social responsibility. This work thus contributes to the promotion of safer academic environments, presenting a system that can also be adapted for other public security applications.