High Fidelity Fingerprint Generation: Quality, Uniqueness, And Privacy

Abstract

In this work, we utilize progressive growth-based Generative Adversarial Networks (GANs) to develop the Clarkson Fingerprint Generator (CFG). We demonstrate that the CFG is capable of generating realistic, high fidelity, 512 × 512 pixels, full, plain impression fingerprints. Our results suggest that the fingerprints generated by the CFG are unique, diverse, and resemble the training dataset in terms of minutiae configuration and quality, while not revealing the underlying identities of the training data. We make the pre-trained CFG model and the synthetically generated dataset publicly available at https: //github.com/keivanB/Clarkson_Finger_Gen

Multiresolution synthetic fingerprint generation

“… Visual comparison between (a) the proposed approach with the PolyU texture, (b) the proposed approach with the Nist SD300a texture, (c) Cao and Jain [22], (d) Finger‐GAN [21], (e) SFinGe [19] and (f) Clarkson Fingerprint Generator (CFG) [24] …”

“…Figure 15 shows a visual comparison between fingerprints generated by the proposed approach, by a publicly available SFinge demo [19], by a public model of Cao and Jain's method [22], by the Clarkson Fingerprint Generator (CFG) method [24], and by our implementation of Finger-GAN [21].…”

“…Visual comparison between (a) the proposed approach with the PolyU texture, (b) the proposed approach with the Nist SD300a texture, (c) Cao and Jain [22], (d) Finger-GAN [21], (e) SFinGe [19] and (f) Clarkson Fingerprint Generator (CFG) [24] F I G U R E 2 7 Quality assessment of the original images, binarised images and the reconstructed binarised images NFIQ…”

“…However, both of these studies present many unnatural artefacts and cannot generate multiple images for the same identity. The Clarkson Fingerprint Generator (CFG) method [24] uses the StyleGAN [25] to synthesise fingerprints. However, the database of real fingerprints captured by the authors is not publicly available.…”

“…F I G U R E 1The Clarkson Fingerprint Generator (CFG) method[24] samples. Note the lack of variation between the different identities (a) and (b), and (c) and (d)1 https://andrewyzy.github.io/MR-SF/.…”