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

2025-10-16

AI Summary

Iris biometrics are vulnerable to presentation attacks (PAs), yet existing presentation attack detection (PAD) methods suffer from scarcity of high-fidelity PA samples. To address this, we propose a multi-domain image translation framework integrating diffusion models and StyleGAN, enabling high-fidelity, diverse synthesis of realistic iris images into various PA modalities—including printed images, artificial eyes, and decorative contact lenses—via latent-space alignment, adaptive perceptual loss, and domain-specific constraints. Our approach is the first to jointly model fine-grained texture details and cross-domain semantic consistency within a unified architecture, substantially improving both realism and generalizability of generated samples. Evaluated on LivDet2020, the synthesized data boosts PAD system performance: the true acceptance rate at 1% false rejection rate increases from 93.41% to 98.72%, demonstrating a significant enhancement in detection robustness and practical utility.

Technology Category

• Computer Vision: Biometrics, Face, Gesture & Pose
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Humans and AI: Interaction Techniques and Devices

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Security and Privacy: Large-scale security measurements

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.

Problem

• Generating synthetic ocular images to address iris presentation attack detection
• Overcoming data scarcity in training iris biometric security systems
• Translating between bonafide and attack domains using multi-domain synthesis

Innovation

• Combines diffusion models with GANs for synthesis
• Translates between bonafide and attack image domains
• Uses adaptive loss for ocular domain consistency