High-Fidelity Fingerprint Image Synthesis via Attention-Enhanced GAN Architecture

Abstract

Fingerprint biometrics is one of the most widely used modalities for secure authentication systems due to its uniqueness and permanence. However, obtaining high-resolution, realistic fingerprint images remains a technical challenge, especially for training robust recognition systems. This research presents a novel approach to fingerprint image synthesis using an attention-enhanced Generative Adversarial Network (GAN) architecture. By integrating spatial and channel attention modules into both the generator and discriminator networks, the proposed model enhances feature representation, resulting in high-fidelity and texturally consistent fingerprint images. Experimental evaluations using public datasets such as FVC2004 and PolyU reveal that the synthesized images closely resemble real samples in terms of quality, ridge structure, and minutiae preservation. Comparative analyses with baseline models such as DCGAN and StyleGAN demonstrate the superiority of the proposed framework in producing visually plausible and biometric-consistent fingerprints. This study contributes a scalable, attention-aware generative model for synthetic fingerprint data generation, beneficial for biometric system development, augmentation, and adversarial robustness testing.