Pixel‐wise supervision for presentation attack detection on identity document cards

Abstract

Identity documents (or IDs) play an important role in verifying the identity of a person with wide applications in banks, travel, video‐identification services and border controls. Replay or photocopied ID cards can be misused to pass ID control in unsupervised scenarios if the liveness of a person is not checked. To detect such presentation attacks on ID card verification process when presented virtually is a critical step for the biometric systems to assure authenticity. In this paper, a pixel‐wise supervision on DenseNet is proposed to detect presentation attacks of the printed and digitally replayed attacks. The authors motivate the approach to use pixel‐wise supervision to leverage minute cues on various artefacts such as moiré patterns and artefacts left by the printers. The baseline benchmark is presented using different handcrafted and deep learning models on a newly constructed in‐house database obtained from an operational system consisting of 886 users with 433 bona fide, 67 print and 366 display attacks. It is demonstrated that the proposed approach achieves better performance compared to handcrafted features and Deep Models with an Equal Error Rate of 2.22% and Bona fide Presentation Classification Error Rate (BPCER) of 1.83% and 1.67% at Attack Presentation Classification Error Rate of 5% and 10%.

Synthetic ID Card Image Generation for Improving Presentation Attack Detection

“…Mudgalgundurao et al [4] proposed a method to detect fake German ID cards and residence permits using pixel-wise supervision based on DenseNet. This technique enables the method to leverage minute cues on various artefacts, such as moiré patterns and artefacts left by the printers.…”

“…Regarding the training and testing data, due to the privacy concerns of the ID cards, the databases for research and commercial purposes are difficult to obtain. So far, there is a limited number of databases available: the Chilean Sequestered ID card database used in [3], Public-IvS where the ID card images are cropped from the existing images from CASIA-IvS [16], and the German ID card and resident permit images [4]. Furthermore, the MIDV-500 dataset [15], which contains video clips of 50 different identity document types, including 17 types of ID cards, 14 types of passports, 13 types of driving licenses is also available.…”

“…After that, a series of algorithms were developed to select random faces and signatures, as well as random names, dates and alphanumeric characters, to fill the templates with random information. The face images are selected from the bona fide probe FERET dataset 3 , while the signatures originated from the Kaggle dataset of handwritten signatures 4 . During the information generation process, we took special care to match the gender of the photo with that of the first name and the gender field in the ID card.…”

“…Even today, there are many templates of ID cards, passports, driver's licenses, and others available to help an attacker produce a high-quality falsified document 2 . Methods that detect fraud in personal ID documents have been presented in recent years [1], [2], [3], [4]. Those systems rely on deep learning methods, such as Convolutional Neural Networks (CNN), in order to achieve great detection accuracy [5].…”

“…https://www.nist.gov/itl/products-and-services/color-feret-database4 https://www.kaggle.com/datasets/divyanshrai/handwritten-signatures perfectly parallel to the document when the user takes pictures of their ID cards . This is replicated using a random protective transformation.…”