Fingerprint and Identity-Marker-Based Server Room Access Security System Using YOLOv8 and Person Re-Identification

Abstract

Server rooms house vital information technology infrastructure, so unauthorized physical access can pose risks of data leakage and operational disruption. This study designs and implements an entry–exit access security system for server rooms that combines fingerprint authentication with three visual identification methods: human detection using YOLOv8n, multi-object tracking with ByteTrack, cross-frame identity continuity maintained through Person ReIdentification (OSNet-AIN), and face verification using ArcFace. When a user's fingerprint and face have both been verified, an ESP32 microcontroller activates a relay and solenoid lock to open the entry door, while at the exit door the user only needs to be re-recognized via Re-ID without repeating full authentication. All access logs are stored in a MySQL database and can be monitored in real time through a FastAPI-based website interface. Testing across variations in distance, light intensity, clothing changes, and simulated tailgating shows that YOLOv8n achieves an F1-Score of 82% at an optimal confidence threshold of 0.439, ArcFace achieves 100% recognition accuracy at an operational distance of 1–2 meters, while OSNet obtains a Rank-1 Accuracy of 31.1% and mAP of 29.1% as a supporting component for body-posture tracking. The system is proven capable of preventing tailgating attempts through validation of entry-log status in the database, demonstrating that a layered visual identification approach combined with IoT-based fingerprint authentication can serve as a more reliable server room security solution than conventional systems.