<p>With the global emergence of stringent privacy protection laws, data collection of crucial nature and its analysis have been impacted. This motivated the research community to introduce and explore areas like data anonymization and the generation of realistic-looking synthetic datasets. Further, with the wide-scale applicability of emerging deep-learning approaches in numerous fields, the public availability of large-scale quality datasets is the need of the hour. In this paper, a high-resolution latent fingerprint dataset is synthetically generated using Generative Adversarial Networks (GANs). The training dataset for the StyleGAN2-ADA architecture has been collected using the Reflected Ultra Violet Imaging System (RUVIS), which works on the principle of reflection of ultraviolet light. The approach assisted in generating anonymous and realistic-looking diverse samples with pore-level details. It will overcome the challenges of the manual collection of large-scale fingerprint datasets. The samples were further annotated using the state-of-the-art crossing number technique. The diversity analysis of synthetic samples was performed using the SSIM and MS-SSIM scores, whereas the CNN model was used for conducting the indistinguishability analysis. A 5-layered CNN model has been used for comparing the accuracy obtained by using real samples and the combination of both the real and the synthetic samples. The accuracy improved from 82.4% to 89.8% when the training dataset contained both synthetic and real data. Further, the generated samples were also visually analyzed by the experts for quality verification.</p>