In this work, we investigate the possibility of generating a grayscale image of the finger vein from its binary template. This exercise would allow us to determine the invertibility of finger vein templates, and this has implications in biometric security and privacy. While such an analysis has been undertaken in the context of face, fingerprint and iris templates, this is the first work involving the finger vein biometric trait. The transformation from binary features to a grayscale image is accomplished using a Pix2Pix Convolutional Neural Network (CNN). The reversibility of 6 different types of binary features is evaluated using this CNN. Further, a number of experiments are conducted using 7 distinct finger vein datasets. Results indicate that (a) it is possible to reconstruct finger vein images from their binary templates; (b) the reconstructed images can be used for biometric recognition purposes; (c) the CNN trained on one dataset can be successfully used for reconstructing images in a different dataset (cross-dataset reconstruction); and (d) the images reconstructed from one set of features can be successfully used to extract a different set of features for biometric recognition (cross-feature-set generalization).

In many real-world prediction tasks, class labels include information about the relative ordering between labels, which is not captured by commonly-used loss functions such as multi-category cross-entropy. Recently, the deep learning community adopted ordinal regression frameworks to take such ordering information into account. Neural networks were equipped with ordinal regression capabilities by transforming ordinal targets into binary classification subtasks. However, this method suffers from inconsistencies among the different binary classifiers. To resolve these inconsistencies, we propose the COnsistent RAnk Logits (CORAL) framework with strong theoretical guarantees for rank-monotonicity and consistent confidence scores. Moreover, the proposed method is architecture-agnostic and can extend arbitrary state-of-the-art deep neural network classifiers for ordinal regression tasks. The empirical evaluation of the proposed rank-consistent method on a range of face-image datasets for age prediction shows a substantial reduction of the prediction error compared to the reference ordinal regression network.

Recent research has established the possibility of deducing soft-biometric attributes such as age, gender and race from an individual's face image with high accuracy. However, this raises privacy concerns, especially when face images collected for biometric recognition purposes are used for attribute analysis without the person's consent. To address this problem, we develop a technique for imparting soft biometric privacy to face images via an image perturbation methodology. The image perturbation is undertaken using a GAN-based Semi-Adversarial Network (SAN) - referred to as PrivacyNet - that modifies an input face image such that it can be used by a face matcher for matching purposes but cannot be reliably used by an attribute classifier. Further, PrivacyNet allows a person to choose specific attributes that have to be obfuscated in the input face images (e.g., age and race), while allowing for other types of attributes to be extracted (e.g., gender). Extensive experiments using multiple face matchers, multiple age/gender/race classifiers, and multiple face datasets demonstrate the generalizability of the proposed multi-attribute privacy enhancing method across multiple face and attribute classifiers.

While the primary purpose for collecting biometric data (such as face images, iris, fingerprints, etc.) is for person recognition, yet recent advances in machine learning has shown the possibility of extracting auxiliary information from biometric data such as age, gender, health attributes, etc. These auxiliary attributes are sometimes referred to as soft biometrics. This automatic extraction of soft biometric attributes can happen without the user's agreement, thereby raising several privacy concerns. In this work, we design a technique that modifies a face image such that its gender as assessed by a gender classifier is perturbed, while its biometric utility as assessed by a face matcher is retained. Given an arbitrary biometric matcher and an attribute classifier, the proposed method systematically perturbs the input image such that the output of the attribute classifier is confounded, while the output of the biometric matcher is not significantly impacted. Experimental analysis convey the efficacy of the scheme in imparting gender privacy to face images.