DRISTHI: AI SECURITY ECOSYSTEM

Contemporary security systems face significant challenges in scalability, cost-effectiveness, and multi-modal threat detection. This paper presents Dristhi, a unified, open-source AI Security Ecosystem for comprehensive, real-time safety monitoring through the seamless integration of five intelligent subsystems: (1) multi angle facial recognition using DeepFace and OpenCV achieving 90.3% accuracy, (2) continuous 24/7 threat surveillance via SmolVLM (mAP@0.5 of 87.6%), (3) emergency gesture recognition through MediaPipe (F1-Score: 91.2%), (4) speech-activated emergency alerting using OpenAI Whisper (WER: 4.8%), and (5) fall detection for elderly individuals powered by YOLOv8 (sensitivity: 93.1%). Built exclusively on open-source technologies, Dristhi achieves system-wide accuracy exceeding 87% with sub-300ms inference latency across all modules. The ecosystem offers a modular, extensible architecture suitable for smart homes, educational institutions, hospitals, and public infrastructure, delivering an estimated 85-90% cost reduction compared to equivalent proprietary systems.