A Lightweight Deep Learning Framework For Fingerprint Liveness Detection

Abstract

Fingerprint Liveness Detection (FLD) is a critical component of biometric authentication systems that protects against presentation attacks using artificial fingerprints fabricated from materials such as silicone, gelatine, and latex. Existing methods based on Convolutional Neural Networks (CNNs) or multimodal biometric traits have shown promising performance; however, they often increase system complexity, computational cost, and hardware requirements. To overcome these limitations, this paper presents a lightweight deep learning framework for robust fingerprint liveness detection.

The proposed system employs an efficient object detection model with an enhanced backbone and a decoupled detection head, enabling the extraction of fine ridge-level features such as pore distribution and distortions, along with global liveness cues including perspiration dynamics and texture irregularities. Unlike multimodal approaches that require auxiliary biometric data, the proposed framework operates solely on fingerprint images, ensuring hardware simplicity while maintaining high discriminative capability.

The model is trained end-to-end on benchmark datasets using advanced regularization techniques and a cosine-annealed Adam optimizer to improve generalization and reduce overfitting. Experimental evaluations demonstrate that the proposed framework achieves superior spoof detection accuracy, strong resistance to novel attack materials, and fast inference speed compared to existing state-of-the-art approaches. With its lightweight architecture and adaptability, the system provides a practical and scalable solution for improving the reliability of biometric authentication in real-world environments.