GRU-AUNet: A Domain Adaptation Framework for Contactless Fingerprint Presentation Attack Detection
Banafsheh Adami, Nima Karimian · 2025
Read GRU-AUNet: A Domain Adaptation Framework for Contactless Fingerprint Presentation Attack Detection by Banafsheh Adami, Nima Karimian (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-01213.txt.
Introduction
Banafsheh Adami and Nima Karimian build a system that tells real touchless fingerprints from fakes. Their 2025 paper addresses a security gap: contactless fingerprint readers are comfortable, but easier to spoof with printed or molded fakes. Their design joins a transformer-based image network with a memory module that sharpens attention, plus a dynamic filter at its core.
Training mixes real and spoof images across three public data sets. Error rates came out very low, under a tenth of a percent for rejecting real users and near one percent for accepting fakes, beating prior domain adaptation methods. The work targets phone-grade convenience without surrendering the door.
Catalog record
| CREATOR | Banafsheh Adami, Nima Karimian |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-01213
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
GRU-AUNet: A Domain Adaptation Framework for Contactless Fingerprint Presentation Attack Detection Banafsheh Adami, Nima Karimian Abstract Although contactless fingerprints offer user comfort, they are more vulnerable to spoofing. The current solution for anti-spoofing in the area of contactless fingerprints relies on domain adaptation learning, limiting their generalization and scalability. To address these limitations, we introduce GRU-AUNet, a domain adaptation approach that integrates a Swin Transformer-based UNet architecture with GRU-enhanced attention mechanisms, a Dynamic Filter Network in the bottleneck, and a combined Focal and Contrastive Loss function. Trained in both genuine and spoof fingerprint images, GRU-AUNet demonstrates robust resilience against presentation attacks, achieving an average BPCER of 0.09% and APCER of 1.2% in the CLARKSON, COLFISPOOF, and IIITD datasets, outperforming state-of-the-art domain adaptation methods. † † address: West Virginia University 1 Introduction Biometric systems have found extensive utility across various domains, including but not limited to law enforcement and forensics, singular identification, healthcare, and facilitating access control for smartphones and tablets. These applications contribute to enhanced convenience in our day-to-day activities. The demand for contactless biometric solutions is increasing rapidly due to hygiene-related issues. Fingerprints and facial biometrics are recognized as the primary modalities in the field of biometrics which extensive implementation by law enforcement agencies and national ID programs on a global scale [ 1 ] . According to the biometric system market is projected to reach a value of $82.9 billion by 2027 [ 2 ] . Despite popularity of face authentication, it has encountered challenges during the pandemic, particularly regarding the use of face coverings [ 3 ] , which prevents its high rate of detection [ 4 , 5 , 6 ] . Contactless fingerprint recognition provides great potential in various applications, offering a touchless and hygienic biometric authentication solution. Contactless fingerprinting is a cutting-edge technological advancement in the field of biometrics that eliminates the need for traditional bioscanner sensors [ 7 , 8 , 9 , 10 , 11 ] . Instead, it relies solely on a smartphone camera lens for capturing and recording fingertip information. Compared to touch based fingerprint, it is considered to be more seamless and convenient and has a higher user acceptance. Figure 1: Different spoofed samples from CLARKSON ,COLFISPOOD, and III-TD datasets were employed in this paper. First four spoof samples are related to CLARKSON dataset (WOODGLUE,PLAYDOH, PHOTOPAPER, ECOFLEX), and others are spoof samples from COLFISPOOF dataset, indoor and outdoor photopaper are realted to III-TD dataset. Although contactless fingerprint technology provides convenient and widely accepted user experiences, it does come with various drawbacks. These include lower biometric performance, susceptibility to environmental influences, and vulnerabilities to presentation attacks [ 12 ] . Presentation attacks can compromise the security and reliability of biometric authentication systems, potentially leading to unauthorized access or identity theft [ 13 ] . Hence, developing an effective countermeasure against contactless fingerprint is crucial to detect and prevent any unseen presentation attacks. Contactless fingerprint systems used smartphones for capturing photo-based finger images are more vulnerable to spoofing due to the use of a single type of camera and limited computational capabilities (Figure. 1 shows fingerprint spoof images fabricated for this study). While well-studied has been conducted on contact-based fingerprint recognition and its vulnerabilities to presentation attacks, limited attention has been given to studying contactless fingerprint presentation attacks. In recent years, only few numbers of approaches have been explored to identify various contactless fingerprint presentation attacks using hand-crafted features, and deeply learned features. Deep learning models have been widely applied in biometrics and other domains to improve feature extraction and classification [ 14 ] . However, despite significant progress in contactless fingerprint PAD, several limitations remain, including: • Disparities in data distributions: existing PAD approaches on contactless fingerprint assume similar data distributions between training and testing scenarios. However, this assumption leads to limited generalization capabilities of PAD methods when faced with real-world situations, especially with unseen attacks. • Multiple types of presentation attacks: contactless fingerprint presentation attacks can take various forms, including printed attack, latex, ecoflex, and more. Thus, generating and creating a labeled training set that covers all possible presentation attacks for each new application scenario is impractical. Our work introduces a novel unsupervised learning approach based on the Swin-UNet architecture with attention mechanisms for contactless fingerprint anti-spoofing. The proposed method demonstrates improved generalization, scalability, and robustness against diverse presentation attacks, as validated through comprehensive experiments on multiple datasets. Our research makes a significant contribution to the field of biometric security by enhancing the security and reliability of contactless fingerprint recognition systems. The main contributions of our work are: • Development of a novel unsupervised learning approach for contactless fingerprint anti-spoofing: We have proposed a new method that combines the Swin-UNet architecture with a Swin Transformer backbone to detect presentation attacks in contactless fingerprint recognition systems. This approach is unsupervised, meaning it is trained solely on genuine fingerprint images without the need for labeled spoof samples during training. • Improved generalization and scalability: By training the model only on genuine fingerprint images, our approach addresses the limitations of existing methods that rely on mixed training data of genuine and spoof samples. This unsupervised learning strategy enhances the model’s ability to generalize to unseen presentation attacks and scales better to new application scenarios. • Comprehensive evaluation: The proposed method has been evaluated on multiple datasets which contains a wide range of presentation attacks, demonstrating its robustness and generalization ability. The comparison with state-of-the-art approaches and the reported performance metrics (APCER, BPCER, and HTER) suggest that the proposed method is effective. • Robustness against various presentation attacks: Our model has been evaluated against a diverse range of presentation attacks, including those from the CLARKSON, COLFISPOOF, and IIITD Spoofed Fingerphoto databases. These databases cover a wide variety of spoofing techniques, such as printed attacks, ecoflex, playdoh, gelatin, and more. Our model’s ability to detect these various attacks demonstrates its robustness and effectiveness. • Integration of attention mechanisms and skip connections: We have modified the Swin-UNet architecture by replacing the skip connections with a novel attention path. This attention path incorporates channel attention and spatial attention to help the model focus on the most relevant features for accurate classification. The integration of attention mechanisms and skip connections enhances the model’s ability to capture discriminative features and improves its overall performance. • Comprehensive evaluation and comparison: We have conducted extensive experiments to evaluate the performance of Our proposed method and compared it with state-of-the-art approaches. The results demonstrate the superiority of our Swin-UNet-based approach in terms of APCER, BPCER, and HTER metrics across multiple datasets. This comprehensive evaluation validates the effectiveness of our unsupervised learning approach for contactless fingerprint anti-spoofing. • Contribution to the field of biometric security: Our work addresses a critical challenge in contactless fingerprint recognition systems, which are increasingly popular due to their convenience and hygienic advantages. By developing an effective anti-spoofing solution that can detect various presentation attacks, our research contributes to enhancing the security and reliability of contactless fingerprint authentication systems, thereby advancing the field of biometric security. 2 Related Work We have summarized the previous works on contactless fingerprint technology in the Table 1 . Fujio et al. [ 15 ] were early pioneers in investigating the use of