Anti-spoofing attendance system using blockchain and AI face recognition

Face recognition systems have become pivotal in modern security frameworks due to their accuracy and convenience. However, these systems are increasingly threatened by presentation attacks, embedding leakage, and tampering risks, especially in sensitive domains such as online examinations and e-governance. To address these challenges, this paper presents a secure and tamper-proof face recognition and attendance system by integrating deep learning-based anti-spoofing techniques with blockchain technology. The proposed system employs SilentFaceNet for real-time spoof detection and InceptionResNetV1 and FaceNet for extracting robust facial embeddings. To ensure privacy and integrity, embeddings are securely stored using IPFS, and attendance events are immutably recorded through Ethereum smart contracts. Experimental evaluation demonstrates that the proposed model achieves high accuracy (97.8%), low ACER (0.7%), and minimal inference time (0.02 s), outperforming traditional methods such as DeepPixBiS and LBP. In addition, the blockchain integration enhances transparency, traceability, and resistance to tampering. The system provides a scalable, explainable, and secure solution for decentralized biometric authentication and trusted attendance management.