Smart lock system with internet of things using facial recognition for home access

Abstract

Security in urban residential environments is a significant challenge in regions with high crime rates and unauthorized access. This has led to the integration of internet of things (IoT) technologies and biometric authentication mechanisms into access control systems. Research has described smart lock solutions that require significant hardware resources or computationally complex recognition algorithms. This research proposes the implementation of a low-cost distributed architecture for facial recognition, lock control, monitoring, and supervision using mobile devices. The proposed architecture performs facial recognition using a histogram-oriented gradient (HOG) model executed on a desktop processing node, while an ESP32 microcontroller activates the lock. Remote monitoring is achieved through a communication framework using Flask libraries, tunneling with Ngrok, and a mobile application. Experimental validation was performed in a controlled environment, achieving 96.2% accuracy, and an F1-score of 95.9%. Furthermore, the control subsystem achieved an activation time of 0.78 seconds. The results validate the integration of biometric authentication, IoT control, and remote monitoring. However, limitations were identified in low light or spoofing environments.