Liveness Detection-Based Home Door Security System for Anti-Spoofing Using Intel RealSense F455 Camera and LBPH

Abstract

Spoofing attacks on facial recognition-based security systems are increasing along with the development of smart home technology. To address this issue, this study proposes a home door security system with the implementation of liveness detection-based anti-spoofing technology using an Intel RealSense F455 camera. The system is designed to verify the authenticity of a user's face in real-time by combining facial texture analysis and the user's physiological responses. The facial detection process is carried out using the Haarcascade algorithm to extract a 160×160 pixel facial area, while facial recognition uses the Local Binary Pattern Histogram (LBPH) method which is relatively stable to variations in lighting and viewing angles. The liveness detection mechanism is implemented mechanically by utilizing the Haarcascade Eye to detect the user's eye movements as an indicator of the presence of a live face, so that the system is able to distinguish real faces from fake media in the form of static photos. The system is integrated with a Telegram bot for real-time access monitoring, where automatic notifications are sent every time a door access attempt occurs. Test results show a facial recognition accuracy rate of 98.93%, with the system successfully detecting and verifying 30 registered users and producing an average confidence value consistently above 80%. Furthermore, the liveness detection mechanism proved effective in preventing photo-based spoofing attacks, with a stable detection success rate throughout the testing. These findings suggest that the integration of LBPH and eye-based liveness detection can improve the reliability of facial recognition-based door security systems.