Security measures like identity verification are extensively implemented and seamlessly integrated into daily life. However, these systems are often not designed to accommodate the needs of neurodivergent individuals, leading to potential accessibility and reliability challenges. Neurodiverse populations exhibit differences in cognitive abilities, behavior, and learning processes, including difficulties with memory and attention. These obstacles prevent users from recalling information or accessing instructions. Furthermore, neurodivergent individuals are more prone to exhibit differences in eye gaze, keystroke dynamics, or speech patterns than neurotypical populations. Individuals with neurodevelopmental conditions often exhibit cognitive and sensory differences that can render traditional identity verification systems less effective. This demographic disparity can increase susceptibility to cyberattacks, underscoring the urgent need for further research. The proposed study highlights these disparities and urges the community to explore strategies to improve fairness and effectiveness in such systems. This paper emphasizes the importance of modifying user identity confirmation technologies to suit better neurodiverse populations and advocates for developing more inclusive and accessible systems. It is crucial to ensure that security technologies, such as biometrics, are inclusive and effective for all users, including neurodivergent users. The proposed research promotes equal access and ethical practices in verifying digital identity.

The rapid advancement of smartphone technology, driven by sophisticated cameras and accelerated computing capabilities, has fueled the demand for reliable and secure identity verification solutions. Finger photo recognition has emerged as a popular and touchless alternative for secure smartphone unlocking. However, systems reliant solely on single RGB images face substantial challenges, including device variability, inconsistent backgrounds, and varying lighting conditions, which can compromise their effectiveness. This research addresses the limitations of current finger-based biometric systems by introducing a novel identity verification system that leverages the rich and dynamic information captured in finger-video data. The proposed approach, VIDVerify, combines binary classification, reduced information redundancy, and loss of cosine embedding to create a robust identity verification pipeline. VIDVerify employs a Siamese architecture with joint embeddings and three complementary loss functions: VICReg loss for redundancy reduction, focal loss for self-class balancing, and cosine embedding loss for video data optimization. The system is evaluated using the Multi-Movement Finger-Video (MMFV) database, which captures finger movements across multiple axes. By comparing convolutional backbones (ResNets, MobileNets) with transformer-based backbones (Swin Transformers), this research establishes a new benchmark for finger-video-based identity verification, paving the way for future advancements in this field. The code is available on GitHub: www.github.com/sulabh-shr/fingerprint.

Hyperspectral imaging (HSI) outperforms the ability of RGB and multi-spectral imaging by conveying information through hundreds of contiguous wavelength intervals. This emerging technology can enable real-time monitoring of spatially resolved spectral information of materials. This paper explores the use of HSI classification to analyze the spectral structure of the human skin of fingers. The spectral reflectance of human skin is believed to vary significantly between individuals, but finding a typical signature for human skin reflectance and what its distribution is across a population are open research questions. The skin spectra acquired through a spectrograph from a diverse population are proven to be interspersed, thus the system using it would not be challenged by ethnicity. Existing related studies are using spectrophotometers only to explore features of human skin reflectance. Rather than use biochemistry, this research enables the design of image-based hand-crafted features to determine its representation. Although the lack of a biochemical component limits identity verification to extrinsic factors, reflectance has the potential as an identifying factor. As the first research to use HSI to assess reflectance, we focus on determining the extent of that potential, including its potential use in machine learning applications for identity verification.

Finger photo recognition has emerged as an alternative biometric authentication solution in smartphones, leveraging common RGB cameras to acquire images of human fingers, improving hygiene and user experience. The security of this technology is currently threatened by presentation attacks. Although being equipped with presentation attack detection modules is of critical importance for these systems, existing approaches are not robust to several challenges including unknown attacks and device diversity. The limited availability of training data to the research community has constrained progress. In this paper, we present two new databases of finger photos for developing anti-spoofing countermeasures, Mason Finger Photo Presentation Attack Detection iPhone 13 Pro 2022 (MFPAD-i-22) and Mason Finger Photo Presentation Attack Detection Google Pixel 32023 (MFPAD-G-23), containing live and spoof finger photos with associated demographics. MFPAD-i-22 was acquired from 112 subjects using the device iPhone 13 Pro, while MFPAD-G-23 from 100 individuals using Google Pixel 3. We also discuss a novel mobile App we developed in an Android environment based on a previously designed PAD fusing different color spaces. By providing these resources and insights, we encourage researchers to spend efforts to advance contactless fingerprint PAD in mobiles, for more secure and robust biometric systems.

A fingerphoto is obtained by imaging a human finger using a basic smartphone camera. Although impressive advances have been made to accurately match fingerphotos, this technology is vulnerable to presentation attacks (PAs). These algorithms do not generalize well in the presence of new presentation attacks. While previous research on this issue is limited, this paper systematically evaluates fingerphoto presentation attack detection (PAD) algorithms under unknown attacks. The proposed assessment compares different Convolutional Neural Networks (CNNs) on the IIITD Smartphone Fingerphoto database with spoof data including printout and various display attacks. These images used for the experiments were acquired indoors and subjected to background (i.e., white or natural) and capture device (i.e., Nokia or OPO) variations. Preliminary results show that the PAD based on AlexNet is robust under most types of replica unseen during the training of the detector.