SafeGate AI: A Real-Time Multi-Stage Content Safety System for Child Protection and Harmful Content Detection on Linux Endpoints

Abstract

SafeGate AI is a real-time, multi-stage content safety system for Linux endpoints that monitors supported OS-level input and file channels to detect potentially harmful content. It combines event monitoring, pattern-based detection, secondary risk scoring, configurable thresholds, enforcement actions, escalation handling, and structured event logging. The paper presents the system architecture, implementation, evaluation, performance results, limitations, and future scope.