Improving fingerprint presentation attack detection by an approach integrated into the personal verification stage

TL;DR

The paper introduces the Closeness Binary Code ($CC$) add-on, a lightweight integration between a standard PAD and fingerprint verification systems that exploits enrolled template information. In Euclidean embedding space, bona fide samples cluster by finger and user, enabling a 3-bit code $\{b_f,b_p,b_o\}$ that informs a LUT-based final decision, without requiring retraining of the PAD. Across white-box and black-box PADs and multiple LivDet datasets, the $CC$ add-on improves BPCER and APCER, enhances integrated AFIS performance in Bio-WISE simulations, and competes with state-of-the-art domain-generalization methods while imposing minimal computational overhead. The approach is practical for personal devices and small-to-medium verification settings, enabling enhanced security by leveraging subject-specific information without data-heavy model retraining.

Abstract

Presentation Attack Detection (PAD) systems are usually designed independently of the fingerprint verification system. While this can be acceptable for use cases where specific user templates are not predetermined, it represents a missed opportunity to enhance security in scenarios where integrating PAD with the fingerprint verification system could significantly leverage users' templates, which are the real target of a potential presentation attack. This does not mean that a PAD should be specifically designed for such users; that would imply the availability of many enrolled users' PAI and, consequently, complexity, time, and cost increase. On the contrary, we propose to equip a basic PAD, designed according to the state of the art, with an innovative add-on module called the Closeness Binary Code (CC) module. The term "closeness" refers to a peculiar property of the bona fide-related features: in an Euclidean feature space, genuine fingerprints tend to cluster in a specific pattern. First, samples from the same finger are close to each other, then samples from other fingers of the same user and finally, samples from fingers of other users. This property is statistically verified in our previous publication, and further confirmed in this paper. It is independent of the user population and the feature set class, which can be handcrafted or deep network-based (embeddings). Therefore, the add-on can be designed without the need for the targeted user samples; moreover, it exploits her/his samples' "closeness" property during the verification stage. Extensive experiments on benchmark datasets and state-of-the-art PAD methods confirm the benefits of the proposed add-on, which can be easily coupled with the main PAD module integrated into the fingerprint verification system.

Figures (9)

• Figure 1: Workflow illustrating the application of the Closeness Binary Code (CC) module in verification scenarios. During enrollment, user fingerprint templates are stored in the database. In the verification stage, the PAD system, enhanced by the CC add-on, processes input samples to classify them as bona fide (BF) or presentation attacks (PA) exploiting template database information. Accepted samples are further compared in the AFIS to verify identity, enabling robust security in small-to-medium-scale setups or personal devices like smartphones.
• Figure 2: Bona fide embeddings from the same user tend to be close in feature space. Knowledge of user-specific information can be used to more accurately classify new user samples into bona-fide or presentation attack, assuming it is supposed to be located within the user cluster in the former case.
• Figure 3: Overview of an AFIS with the Closeness binary Code add-on. The system combines the claimed identity templates with the validation set $V$ to compute the closeness sequence. The corresponding outcome from the Look-Up Table (LUT) is then integrated with the AFIS decision to derive the final authorization status.
• Figure 4: Experimental protocol to evaluate the effectiveness of using the CC module in increasing the performance of a PAD. The test set samples are classified both by the pre-trained model and by the CC method, which uses not only the validation set V but also the bona-fide samples of the same user as a template gallery for calculating the closeness binary code.
• Figure 5: t-SNE visualization of feature embeddings for a random user from GreenBit 2021 dataset using the CNN PADs. Each color represent a different finger of the user.