Transplant Track: Privacy-Preserving Smart Attendance System with Face and Liveness Detection

The conventional way of keeping track of attendance in learning institutions is through manual roll calls or physical biometric systems, which are not effective and could also pose security challenges in relationships with privacy. The study offers a smart attendance system that protects privacy through face recognition and face detection that are applied to edge devices. The system uses computer vision and deep learning technologies including the OpenCV, DeepFace, Dlib, MTCNN and FaceNet to identify and determine student faces in real time. Eye blink analysis is implemented to detect passive liveness in order to decrease spoofing attacks via photos or videos. Embeddings on the face are stored in an encrypted SQLite database with user identifiers being hashed with a hash (SHA-256) to improve the security of privacy. The attendance records are recorded on the local disk with time stamps and have been saved as an append only file (csv). The system does not require cloud storage and the risk of privacy is less because all data processing is done in the device, which is constructive. It is proven through experimental assessment that the system is highly recognition accurate with low latency, and thus can be used in classroom settings.