Privacy-Preserving Face Recognition for Smart Locks using TensorFlow Lite and BLE

This paper presents the design and development of an automatic door locking system based on Bluetooth with face recognition for secure, contactless access control. The system utilizes a mobile application built using Flutter that allows users to remotely control and monitor access. Facial recognition is done through a light-weight neural network model executed using TensorFlow Lite for privacy and efficiency in computation on mobile devices themselves. The ESP32 microcontroller also supports Bluetooth Low Energy (BLE) communication that enables wireless communication with the door lock mechanism using low power with high reliability. Experimental testing of the system also demonstrates its operation under various conditions of light, Bluetooth distances, and usage frequency, with exact high rates, low response time, and maximization of power consumption. The solution proposed is an effective, cost-effective, and scalable solution compared to traditional keypad or biometric systems and can be implemented in both home and commercial environments.