Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Multi-domain Image Translative Diffusion StyleGAN for Iris Presentation Attack Detection

Shivangi Yadav, Arun Ross

TL;DR

This work addresses the scarcity of iris PAD data by introducing MID-StyleGAN, a diffusion-assisted StyleGAN architecture that performs multi-domain image translation across bonafide, printed-eye, and cosmetic contact lens domains. By integrating an encoder, a domain‑aware, diffusion‑timstep discriminator, and domain transfer losses, the method generates realistic, high-resolution ocular images and enables effective synthetic data augmentation for PAD. Empirical results show superior realism (average FID ≈ 19.71) compared to baselines and clear PAD gains in cross-dataset tests, with notable improvements when synthetic data are used for training detectors. The approach offers a scalable solution to data scarcity in iris biometrics, while acknowledging privacy considerations and suggesting future work to mitigate potential identity leakage.

Abstract

An iris biometric system can be compromised by presentation attacks (PAs) where artifacts such as artificial eyes, printed eye images, or cosmetic contact lenses are presented to the system. To counteract this, several presentation attack detection (PAD) methods have been developed. However, there is a scarcity of datasets for training and evaluating iris PAD techniques due to the implicit difficulties in constructing and imaging PAs. To address this, we introduce the Multi-domain Image Translative Diffusion StyleGAN (MID-StyleGAN), a new framework for generating synthetic ocular images that captures the PA and bonafide characteristics in multiple domains such as bonafide, printed eyes and cosmetic contact lens. MID-StyleGAN combines the strengths of diffusion models and generative adversarial networks (GANs) to produce realistic and diverse synthetic data. Our approach utilizes a multi-domain architecture that enables the translation between bonafide ocular images and different PA domains. The model employs an adaptive loss function tailored for ocular data to maintain domain consistency. Extensive experiments demonstrate that MID-StyleGAN outperforms existing methods in generating high-quality synthetic ocular images. The generated data was used to significantly enhance the performance of PAD systems, providing a scalable solution to the data scarcity problem in iris and ocular biometrics. For example, on the LivDet2020 dataset, the true detect rate at 1% false detect rate improved from 93.41% to 98.72%, showcasing the impact of the proposed method.

A Multi-domain Image Translative Diffusion StyleGAN for Iris Presentation Attack Detection

TL;DR

This work addresses the scarcity of iris PAD data by introducing MID-StyleGAN, a diffusion-assisted StyleGAN architecture that performs multi-domain image translation across bonafide, printed-eye, and cosmetic contact lens domains. By integrating an encoder, a domain‑aware, diffusion‑timstep discriminator, and domain transfer losses, the method generates realistic, high-resolution ocular images and enables effective synthetic data augmentation for PAD. Empirical results show superior realism (average FID ≈ 19.71) compared to baselines and clear PAD gains in cross-dataset tests, with notable improvements when synthetic data are used for training detectors. The approach offers a scalable solution to data scarcity in iris biometrics, while acknowledging privacy considerations and suggesting future work to mitigate potential identity leakage.

Abstract

An iris biometric system can be compromised by presentation attacks (PAs) where artifacts such as artificial eyes, printed eye images, or cosmetic contact lenses are presented to the system. To counteract this, several presentation attack detection (PAD) methods have been developed. However, there is a scarcity of datasets for training and evaluating iris PAD techniques due to the implicit difficulties in constructing and imaging PAs. To address this, we introduce the Multi-domain Image Translative Diffusion StyleGAN (MID-StyleGAN), a new framework for generating synthetic ocular images that captures the PA and bonafide characteristics in multiple domains such as bonafide, printed eyes and cosmetic contact lens. MID-StyleGAN combines the strengths of diffusion models and generative adversarial networks (GANs) to produce realistic and diverse synthetic data. Our approach utilizes a multi-domain architecture that enables the translation between bonafide ocular images and different PA domains. The model employs an adaptive loss function tailored for ocular data to maintain domain consistency. Extensive experiments demonstrate that MID-StyleGAN outperforms existing methods in generating high-quality synthetic ocular images. The generated data was used to significantly enhance the performance of PAD systems, providing a scalable solution to the data scarcity problem in iris and ocular biometrics. For example, on the LivDet2020 dataset, the true detect rate at 1% false detect rate improved from 93.41% to 98.72%, showcasing the impact of the proposed method.
Paper Structure (15 sections, 11 equations, 5 figures, 4 tables)

This paper contains 15 sections, 11 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Generative Adversarial Networks (GANs)
  4. Diffusion-GANs
  5. Proposed Method
  6. Experiments and Analysis
  7. Datasets and PA Detection Methods
  8. Realism Assessment
  9. Utility of Generated Dataset
  10. Baseline Experiment-0
  11. Utility Experiment-1
  12. Ablation Study
  13. Studying Components of Proposed Method
  14. Successive Training Using Synthetic Data
  15. Conclusion and Future Work

Figures (5)

  • Figure 1: Examples of bonafide and different types of iris presentation attacks (PAs). (a) Bonafide, (b) printed eye, (c) artificial eye, and (d) cosmetic contact lens. These images are taken from Berc-IrisFake Lee2007, CASIA-IrisFake Sun2014 and LivDet-2017 Yambay2017 datasets.
  • Figure 2: Illustration of the proposed method that has three modules: (1) Encoder, $E$, which takes an image and its domain label as an input and outputs the encoded image, (2) $G$ that takes the encoded image as an input along with the target domain label to which the input image has to be translated, and (3) $D$ that takes an image and its label as input, and outputs the image probability of domains as well as whether the image is real or synthetic.
  • Figure 3: Samples of ocular images generated using proposed method, MID-StyleGAN. The proposed method is capable of not only generating images from multiple PA domains but also capture intra class variations present in different types of PAs.
  • Figure 4: Comparison of FID scores across multiple generative methods with respect to proposed method. The first plot shows performance across all methods, while the second focuses on realism of images generated using MID-StyleGAN across different domains.
  • Figure 5: Performance of iris PA detectors when trained using only real images and also when trained using real+synthetic images showcasing the usefulness of the generated ocular PA dataset.