Nine Years of Pediatric Iris Recognition: Evidence for Biometric Permanence

Biometric permanence in pediatric populations remains poorly understood despite widespread deployment of iris recognition for children in national identity programs such as India’s Aadhaar and trusted traveler programs like Canada’s NEXUS. This study presents a comprehensive longitudinal evaluation of pediatric iris recognition, analyzing 276 subjects enrolled between ages 4-12 and followed up to nine years through adolescence. Using 18,318 near-infrared iris images acquired semi-annually, we evaluated commercial (VeriEye) and open-source (OpenIris) systems through linear mixed-effects models that disentangle enrollment age, developmental maturation, and elapsed time while controlling for image quality and physiological factors. False non-match rates remained below 0.5% across the nine-year period for both matchers using pediatric-calibrated thresholds, approaching adult-level performance. However, we reveal significant algorithm-dependent temporal behaviors: VeriEye’s apparent decline reflects developmental confounding across enrollment cohorts rather than genuine template aging, while OpenIris exhibits modest but genuine temporal aging (0.5 standard deviations over eight years). Image quality and pupil dilation constancy dominated longitudinal performance, with dilation effects reaching 3.0-3.5 standard deviations, substantially exceeding temporal factors. Failures concentrated in 9.4% of subjects with persistent acquisition challenges rather than accumulating with elapsed time, confirming acquisition conditions as the primary limitation. These findings justify extending conservative re-enrollment policies, potentially to 10-12 year validity periods for high-quality enrollments at ages 7+, and demonstrate iris recognition remains viable throughout childhood and adolescence with proper imaging control.

💡 Research Summary

This paper presents the most extensive longitudinal evaluation of pediatric iris recognition to date, addressing the critical question of biometric permanence in children—a topic that has received little empirical attention despite the widespread deployment of iris-based identity schemes such as India’s Aadhaar and Canada’s NEXUS. The authors enrolled 276 participants aged 4–12 years and followed them for up to nine years, collecting a total of 18,318 near‑infrared iris images in semi‑annual sessions using a single, ISO‑compliant sensor (IrisGuard IG‑AD100). After a brief COVID‑19‑related interruption, 14 collection points were completed, yielding a median follow‑up of 6.5 years and a maximum age span of 4–19 years for the cohort.

Two fundamentally different matchers were evaluated on the identical image set: the commercial VeriEye SDK (version 12.4) and the open‑source OpenIris (v2.0). VeriEye follows a proprietary pipeline with automatic segmentation, quality scoring, and a similarity score where higher values indicate a match. OpenIris implements Daugman’s iris‑code approach augmented by deep‑learning segmentation, producing binary codes compared via fractional Hamming distance (lower values indicate a match). Because adult‑calibrated decision thresholds (VeriEye = 36, OpenIris = 0.35 HD) lead to unacceptably high false‑non‑match rates (FNMR) in children, the authors empirically re‑calibrated thresholds on the full pediatric dataset to achieve a target false‑match rate (FMR) of ≈0.1 % while minimizing FNMR. The resulting operating points were VeriEye = 34 (FMR = 0.11 %, FNMR = 0.16 %) and OpenIris = 0.42 HD (FMR = 0.10 %, FNMR = 0.29 %).

Performance was measured using a fixed‑gallery protocol: each subject’s first‑session images formed the enrollment template, and all subsequent images served as probes. This design isolates true template aging from session‑to‑session variability. Genuine comparisons were paired with a large set of impostor comparisons to estimate FMR. The authors also extracted two key covariates for each image: a composite quality score (0–100) derived from VeriEye’s ISO‑based metric, and a pupil‑dilation ratio (pupil radius / iris radius). Dilation constancy (DC = 1 – |D_gallery – D_probe|) quantified how similar the dilation states were between enrollment and verification.

To disentangle the intertwined effects of elapsed time, biological age, and cohort (enrollment‑age) effects, the authors employed linear mixed‑effects models with an age‑period‑cohort (APC) parameterization. Fixed effects included elapsed time, current age, enrollment age, image quality (gallery and probe), and DC; random intercepts captured subject‑specific variability. This statistical framework allowed the authors to separate genuine template aging from developmental confounding—a challenge that has limited prior longitudinal biometric studies.

Key findings: (1) Both matchers maintained FNMR below 0.5 % across the entire nine‑year horizon when operating at the pediatric‑calibrated thresholds, approaching adult‑level reliability. (2) VeriEye exhibited an apparent decline in genuine scores over time; however, APC modeling revealed that this trend is fully explained by cohort effects—older enrollment cohorts naturally have higher baseline scores, and the observed “aging” is a statistical artifact rather than true template degradation. (3) OpenIris showed a modest but statistically significant genuine aging effect of roughly 0.5 standard deviations over eight years, indicating that algorithmic sensitivity to subtle physiological changes (e.g., pupil dilation dynamics) can manifest as measurable performance drift. (4) Image quality and dilation constancy dominated longitudinal performance: each standard‑deviation improvement in quality or DC contributed 3.0–3.5 SDs of performance gain, dwarfing the pure temporal aging effect (≤0.5 SD). (5) Approximately 9.4 % of subjects experienced persistent acquisition failures (motion blur, eyelid/eyelash occlusion, off‑angle gaze) that were independent of elapsed time, underscoring that high‑quality initial capture is the primary bottleneck for long‑term success.

Operational implications are substantial. The authors argue that, for enrollments performed at age ≥ 7 with high‑quality images and controlled dilation, re‑enrollment intervals can be safely extended to 10–12 years without compromising security (FMR ≈ 0.1 %) or usability (FNMR < 0.5 %). This recommendation directly challenges current policies (e.g., Aadhaar’s mandatory re‑enrollment at ages 5 and 15, NEXUS’s five‑year renewal) that lack longitudinal evidence. Moreover, the algorithm‑dependent aging behavior suggests that system designers should prioritize matchers that are less sensitive to physiological variability (e.g., VeriEye) when deploying iris biometrics for children, or alternatively invest in robust dilation‑normalization techniques if using open‑source pipelines.

In summary, the study provides compelling empirical proof that the iris texture remains remarkably stable throughout childhood and adolescence, provided that imaging conditions—especially image quality and pupil‑dilation consistency—are rigorously managed. The sophisticated APC mixed‑effects analysis convincingly separates true biometric aging from developmental confounding, delivering a nuanced understanding of how different recognition algorithms behave over long time spans in a pediatric population. These insights enable evidence‑based policy revisions, cost‑effective system design, and broader confidence in the use of iris biometrics for inclusive, lifelong identity management.