Generative Adversarial Networks Applied for Privacy Preservation in Biometric-Based Authentication and Identification

2025-09-24

AI Summary

To address the uncontrollable privacy risks, leakage, and misuse of biometric data in biometric authentication, this paper proposes a visual-domain-transfer privacy-preserving method based on Generative Adversarial Networks (GANs). Specifically, it maps facial images across domains into a semantically unrelated private domain (e.g., flowers or shoes), rendering the original biometric features neither visually discernible nor reconstructible via inversion. An identity classifier is jointly trained on the private-domain images to preserve authentication utility while ensuring privacy security. Extensive experiments demonstrate that the proposed approach maintains robustness against diverse reconstruction attacks, membership inference, and adversarial attacks. Its authentication accuracy closely approaches that of the original domain baseline and significantly outperforms existing privacy-preserving methods. Crucially, the method achieves a substantive trade-off between strong privacy guarantees and practical usability.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Privacy
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Biometric-based authentication systems are getting broadly adopted in many areas. However, these systems do not allow participating users to influence the way their data is used. Furthermore, the data may leak and can be misused without the users' knowledge. In this paper, we propose a new authentication method that preserves the privacy of individuals and is based on a generative adversarial network (GAN). Concretely, we suggest using the GAN for translating images of faces to a visually private domain (e.g., flowers or shoes). Classifiers, which are used for authentication purposes, are then trained on the images from the visually private domain. Based on our experiments, the method is robust against attacks and still provides meaningful utility.

Problem

• Biometric authentication systems lack user control over data usage
• Biometric data risks unauthorized leakage and potential misuse
• Need privacy-preserving authentication maintaining security and utility

Innovation

• GAN transforms face images to private domain
• Classifiers trained on visually private domain images
• Method ensures privacy while maintaining authentication utility