Efficient Employee Attendance System Integrating RFID and Android-Based Face Recognition with Liveness Detection

Attendance systems are crucial in corporate management, as accurate attendance recording allows companies to manage human resources more efficiently, identify attendance patterns, and improve employee productivity. With the advancement of information technology, attendance systems have utilized various technologies, such as RFID cards, biometrics like fingerprints or facial recognition, and mobile-based attendance systems. RFID-based attendance systems are widely used due to their ease of quickly and accurately recording attendance. However, these systems have limitations, such as the possibility of RFID cards being lost or damaged and being lent or misused by others, thus not always ensuring the actual physical presence of employees. Therefore, integrating attendance systems with RFID and human biometric features is necessary to ensure employee presence. This study proposes an Android-based attendance system using face recognition and liveness detection technology with CNN (Convolutional Neural Network) integrated with RFID. This technology ensures attendance is recorded only if the employee is on-site and recognized as a genuine cardholder. The MiniFASNet model, which used Fourier transforms to transform the face image into a frequency wave signal as a liveness detection feature, was combined with the FaceNet model for facial feature extraction. Cosine similarity is then used to determine the similarity between real-time captured facial features and the stored dataset features. The experiment results showed a liveness detection confidence score of 0.9 and an average cosine similarity threshold of 0.65 for face recognition. The best recognition accuracy achieved was 98.46% at a distance 50 cm with a front-facing pose and bright lighting conditions.