Threat Sense-XAI: A Context-Aware Threat Intelligence Framework for Real-Time Weapon, Violence, and Behavioural Risk Assessment

Abstract

Classic approaches to security monitoring often encounter problems when dealing with detecting and resolving security threats in real time. This paper introduces a system that uses context-aware threat intelligence to detect weapons, recognize violent actions, analyze time-based patterns, and assess potential threats based on the situation, all to spot unusual behaviour in real-time video feeds. The proposed system utilizes YOLO-based deep learning models to detect weapons and classify human behaviour. Alternatively, a scoring engine determines the level of risks from the objects observed, the behaviour pattern, the location, and past occurrences. High-risk occurrences will automatically be accompanied by raising an alarm, collecting the evidence, reporting the occurrence, and sending emails. In order to foster openness in the process and increase the level of confidence, Explainable Artificial Intelligence (XAI) methods including SHAP (SHapley Additive exPlanations) and Grad-CAM are applied to enhance transparency. Also, a real-time dashboard for incidents is put up for monitoring and analysing incidents.