In this work, we examine if current state-of-the-art deep learning face recognition systems exhibit a negative bias (i.e., poorer performance) for children when compared to the performance obtained on adults. The systems selected for this work are five top performing commercial-off-the-shelf face recognition systems, two government-off-the-shelf face recognition systems and one open-source face recognition solution. The datasets used to evaluate the performance of the systems are both unconstrained in age, pose, illumination, and expression and are publicly available. These datasets are indicative of photo journalistic face datasets published and evaluated on over the last few years. Our findings show a negative bias for each algorithm on children. Genuine and imposter distributions highlight the performance bias between the datasets further supporting the need for a deeper investigation into algorithm bias as a function of age cohorts. To combat the performance decline on the child demographic, several score-level fusion strategies were evaluated. This work identifies the best score-level fusion technique for the child demographic.

/pdf/face-recognition-algorithm-bias-performance-differences-on-4u80jfwn23.pdf

Face Recognition Algorithm Bias: Performance Differences on Images of Children and Adults

We present a longitudinal study of face recognition performance on Children Longitudinal Face (CLF) dataset containing 3,682 face images of 919 subjects, in the age group [2,18] years. Each subject has at least four face images acquired over a time span of up to six years. Face comparison scores are obtained from (i) a state-of-the-art COTS matcher (COTS-A), (ii) an open-source matcher (FaceNet), and (iii) a simple sum fusion of scores obtained from COTSA and FaceNet matchers. To improve the performance of the open-source FaceNet matcher for child face recognition, we were able to fine-tune it on an independent training set of 3,294 face images of 1,119 children in the age group [3,18] years. Multilevel statistical models are fit to genuine comparison scores from the CLF dataset to determine the decrease in face recognition accuracy over time. Additionally, we analyze both the verification and open-set identification accuracies in order to evaluate state-of-the-art face recognition technology for tracing and identifying children lost at a young age as victims of child trafficking or abduction.

Longitudinal Study of Child Face Recognition

The widespread deployment of multi-biometrics to authenticate users prompts the need for biometric systems with high recognition performance. Further, the biometric data, once leaked or stolen, remains compromised forever. Hence biometric security is of utmost importance. Existing biometric template protection schemes either degrade the recognition performance or they have issues with security and speed. We propose a cancelable multi-biometric authentication approach where a novel bit-wise encryption scheme transforms a biometric template to a protected template using a secret key generated from another biometric template. It fully preserves the number of bit-errors in the original and the protected template to ensure recognition performance equivalent to the performance of the unprotected systems. We introduce Algorithm I and Algorithm II for bit-wise encryption; both are defined over cryptographic-primitives- block cipher based encryption and keyed-hash function. We profile these algorithms on various hardware architectures to calculate the efficiency in terms of the time taken during enrolment and authentication phase. For Algorithm II , we observe that a 3.3 GHz desktop architecture takes about 18 milliseconds on an average of over 200 runs to authenticate a user. Additionally, we provide mathematical proof to show that the proposed scheme guarantees secrecy and irreversibility. The results of comparisons with the existing biometric template protection schemes on the various face and iris databases show that the proposed work provides significantly good recognition performance and efficiency, while it achieves high security. Finally, the bit-wise encryption scheme can be built over the commercial-off-the-shelf systems to achieve security with equivalent high performance.

Cancelable Multi-Biometric Approach Using Fuzzy Extractor and Novel Bit-Wise Encryption

https://cris.vub.be/ws/files/69365659/1_s2.0_S0262885621000147_main.pdf

A survey of iris datasets

Biometric recognition is often used for adults for a variety of purposes where an individual’s identity must be ascertained. However, the biometric recognition of children is an unsolved challenge. Solving this challenge could protect children from identity theft and identity fraud, help in reuniting lost children with their parents, improve border control systems in combatting child trafficking, and assist in electronic record-keeping systems. In order to begin the development of biometric recognition systems for children, researchers collected fingerprint, iris, and outer ear shape biometric information from infants. Each modality provides different challenges. Where possible, the performance of existing hardware and software that was developed for adults was assessed with infants. Where necessary, novel hardware or software was developed. For the ear modality, existing hardware and software which have previously been applied to adults were applied to children. For the iris modality, existing hardware was used to acquire the images, while adjustments to the existing preprocessing algorithms were applied to cater for the localisation and segmentation of infant irises. For the fingerprint modality, novel hardware and image processing software were developed to acquire fingerprints from infants, and convert the images into a format which is backward compatible with existing international standards for minutiae extraction and comparison. The advantages and disadvantages of using each of these modalities during the first year of life were compared, based on both qualitative assessments of usage, and quantitative assessments of performance. While there is no conclusively best modality, recommendations of usage for each modality were provided.

/pdf/biometric-recognition-of-infants-using-fingerprint-iris-and-3rnw5w7des.pdf

Biometric Recognition of Infants using Fingerprint, Iris, and Ear Biometrics

In many applications such as law enforcement, attendance systems, and medical services, biometrics is utilized for identifying individuals. However, current systems, in general, do not enroll all possible age groups, particularly, toddlers and pre-school children. This research is the first of its kind attempt to prepare a multimodal biometric database for such potential users of biometric systems. In the proposed database, face, fingerprint, and iris modalities of over 100 children (age range of 18 months to 4 years) are captured in two different sessions, months apart. We also perform benchmarking evaluation of existing tools and algorithms to establish the baseline results for different unimodal and multimodal scenarios. Our experience and results suggest that while iris is highly accurate, it requires constant adult supervision to attain cooperation from children. On the other hand, face is the most easy-to-capture modality but yields very low verification performance. We assert that the availability of this database can instigate research in this important research problem.

Pratichi Basak

Papers

Multimodal biometric recognition for toddlers and pre-school children