Enhancing Biometric Security: Distributed Data Parallel Acceleration for Generative Adversarial Networks in Synthetic Fingerprint Generation

Abstract

This paper presents a novel approach to fingerprint generation using Generative Adversarial Networks (GANs) optimized through Distributed Data Parallel (DDP) acceleration. Fingerprint generation is a critical task in biometric security systems, and traditional methods often suffer from inefficiencies and scalability issues when handling large datasets or complex models. Our proposed method leverages the power of DDP to distribute the computational load across multiple devices, significantly reducing training time and improving the quality of generated fingerprints. We conduct extensive experiments to demonstrate the effectiveness of our approach, showing that it not only enhances the generation process but also maintains high accuracy and diversity in the generated fingerprints. This method opens up new possibilities for scalable and efficient biometric data generation, which can be integrated into real-world applications with high computational demands.