SECURE EYE: An Intelligent Real-Time Surveillance System for Cyber Threat Detection and Monitoring

Abstract

Conventional campus CCTV systems operate passively and remain vulnerable to presentation attacks, including printed images and digital replay. Although recent studies have advanced human detection, face recognition, and liveness verification, these components are typically evaluated independently, with limited validation of their integration under resource-constrained, CPU-only deployment environment. This study proposes SECURE EYE, an integrated real-time surveillance framework combining YOLOv8n-based human detection, SFace embedding with Support Vector Machine (SVM) classification, and rule-based liveness verification. The system was evaluated on a dataset of 1,050 facial images from 10 identities (80:20 split) and further assessed using 1080p RTSP video streams on a virtual machine (8 vCPU, 8 GB RAM, no GPU). The results show that the system achieves 84.29% face recognition accuracy , with F1-scores ranging from 0.75 to 0.93. The liveness detection module yields a True Positive Rate (TPR) of 80.00% and a True Negative Rate (TNR) of 87.50%, with a False Acceptance Rate (FAR) of 12.50% and a False Rejection Rate (FRR) of 20.00%. Real-time performance reaches 4.77 FPS with 0.21-second latency and moderate CPU utilization (16.89%). These findings demonstrate the feasibility of integrated surveillance with anti-spoofing capabilities in CPU-only deployment environment. However, the FRR of 20% indicates a significant usability limitation, highlighting the need for further optimization.