Effect of pupil dilation on biometric iris recognition systems for personal authentication

Abstract

Purpose:

To study the effect of pupil dilation on a biometric iris recognition (BIR) system for personal authentication and identification.

Methods:

A prospective, non-randomized, single-center cohort study was conducted on patients who reported for a routine eye check-up from November 2017 to November 2019 (2 years). An iris scanning device “IRITECH-MK2120U” was used to initially enroll the undilated eyes. Baseline scans were taken after matching with the enrolled database. All eyes were topically dilated and matched again with the enrolled database. The Hamming distance (a measure of disagreement between two iris codes) and recognition status were recorded from the device output, and eyes were evaluated by slit-lamp ophthalmoscopy with special emphasis on pupil shape, size, and texture.

Results:

All 321 enrolled eyes matched after topical dilation. The pupil size had a significant effect on Hamming distance with a P value <0.05. There were no false matches. A correct recognition rate of 100% was obtained after dilation. No loss of iris texture or pupil shape was observed after dilation.

Conclusion:

A BIR system is a reliable method for identification and personal authentication after pupil dilation. Topically dilated pupils are not a cause for non-recognition of iris scans.

An accurate automated personal authentication is a challenging problem. Biometrics-based authentication systems are automated security technologies that use unique and identifiable physiological and behavioral characteristics of individuals.[1,2]

Out of various physiological and behavioral characteristics, such as assigned natures, used to identify individuals, iris recognition is considered one of the most reliable forms of automatic human identity verification. These Iris recognition systems use iris textures as unique identifiers. The human iris has an intricate structure with various minute characteristics such as furrows, freckles, crypts, and coronas.[3,4] These characteristics are unique because of individual differences that arise during embryonic development. The detailed fine structure of an iris is not genetically determined but develops by a random process. Therefore, the iris patterns of the two eyes of a person or those of identical twins are completely independent and uncorrelated.[5]

Biometric iris recognition (BIR) is one of the most useful recently adopted technologies for security checks and plays a crucial role in a wide range of applications in personal identification. The reasons for using BIR are based on multiple advantages as compared to other biometric systems like fingerprints, retinal scans, voice typing, signature analysis, DNA analysis, and gait and odor analysis. It is a non-contact, fast-processing, and real-time measurement method for recognizing a person’s identity by mathematical analysis of the random patterns that are visible and constant within the iris structure of the individual eye. Also, with the ongoing COVID-19 pandemic, the use of face masks and no contact procedures are being implied along with social distancing in day-to-day life to restrict the spread of this pestilential and enduring virus, which increases the need of the hour to adapt modalities that can be executed at an approving distance. Older methods of authentication like fingerprint matching require contact of masses with the scanning modality, unlike face recognition and iris scanning which can be carried out at a safe distance from the operating device and the person. The iris has been described as an ideal structure of the human body for biometric identification because of its well-protected status against wear and tear, immutable nature, and uniqueness over a wide area of the population including genetically similar twins.[5,6] The formation of the iris begins during the third month of gestation.[7] The patterns are developed by the eighth month,[8] although pigment accretion may continue in the first postnatal year.

Most of the studies conducted in the past were performed outside India and included lesser number of cases as compared to our study. Therefore, the present study was conducted on a large number of cases from the Indian population. Some of the studies conducted in the past have shown that a variation in pupil size decreases the iris biometric performance by increasing the likelihood of false non-matches or rejections and that medication may deform the iris differently.[6,9,10]

Few studies in the past have reported that after pupil dilatation, there may be a mismatch by iris recognition-based biometric systems.

Because the unique identification project of India also known as Aadhaar is based on iris recognition, we want to find out whether after dilation the patient can be identified by an iris recognition-based biometric system or not because pupil dilatation is the easiest way to be utilized by an individual to escape iris recognition-based biometric systems. Also, most of the studies conducted in the past were performed and included a lesser number of cases as compared to our study. 1% tropicamide and 2.5% phenylephrine are typically used mydriatic agents that have a different impact on iris muscles. It has been reported that 1% tropicamide is more effective at inducing pupil dilation than 2.5% phenylephrine.[11] With the increasing prospects of this method of biometric recognition system in various fields, it becomes essential to evaluate its effectiveness and reliability. The Unique Identification Authority of India (UIDAI) is the world’s biggest multi-modal biometric system that uses fingerprint, face, and iris biometric accounts. India’s Unique Identification project is also known as Aadhaar, a word that means “the foundation” in several Indian languages.[12] The Aadhaar program aims toward providing reliable national ID documents to most of estimated India’s 1.25 billion inhabitants. This plan was launched in 2009 with the purpose of eliminating duplicate and sham identities and effortless verification and authentication of individual identities.[13] The technological advancements in security systems have led criminals to resort to various refined measures to escape these systems. However, iris recognition systems become imperceptibly weak when it comes to images that are captured from images or videos that are of artificial iris images, images in which contact lenses are worn by the individual, and the images generated by the computer.[14] Other studies have also highlighted other challenges such as occlusion due to eyelashes and eyelids, reflection due to wearing eyeglasses rotations or off-angle, iris deformities due to diseases, and high spectrum illumination.[15] Topical dilation of the pupil is a routine practice before and after cataract surgery, for retinal examinations and laser procedures, and in various eye pathologies like uveitis, but it can also be used for self-administration which can be used to conquer iris-based biometric systems for various reasons.

With the hypothesis that pupil dilation can distort the iris structure and lead to non-recognition of the previously enrolled eyes in the biometric system, the present study was carried out to study the effect of pupil dilation on an automatic BIR system. These doubts raise a query on the reliability of this scanning modality and also the need for ophthalmologists to increase their awareness of whether there is any need for certification by the treating ophthalmologist for patients who underwent an iris dilation process and would be undergoing any biometric iris scanning process for identification processes like at airports or a workplace until the effect of the drug lasts.

Methods

Setting

The present study was a hospital-based prospective, non-randomized, single-center cohort study conducted on a single eye of 321 patients who reported for routine eye check-up from November 2017 to November 2019 (2 years) at a tertiary healthcare center in India for 2 years, for the evaluation of the performance of BIR systems after pupil dilation with tropicamide 0.8% eye drops and phenylephrine 5% eye drops. Approval was obtained from the institutional ethics committee. Patients having nystagmus, very young patients, uncooperative patients, patients having dense corneal opacities and hazy media obstructing the view of the iris, patients having a pre-existing irregular pupil due to posterior synechiae, and patients having active infective pathologies of the eye or surrounding skin, a nondilating pupil, ring synechiae, total posterior synechiae, a pre-existing dilated pupil (post-trauma/third nerve palsy), and iris atrophy were excluded from the study. Nondilating pupils were excluded, and poorly dilated pupils were not excluded from the study.

Consent

The study adhered to the tenets of the Declaration of Helsinki. Well-informed written consent for ophthalmic examination and repeated automatic biometric iris scans was obtained from eligible patients participating in the study, and they were assured that this was only for the purpose of academic study and that their personal and biometric details will remain confidential and will not be made public.

Iris scanning process

All patients underwent the enrollment process after iris scans through a government-certified iris scanning machine, and data provided by the scanner (Hamming distance and recognition status) were recorded after every identification process. The Hamming distance is the mathematical difference between two iris codes and recognition status: iris codes match when the Hamming distance between identification and enrollment iris codes is below the system threshold value. For capturing the iris images, we used the device “IRITECH-MK2120U” which is an ultra-compact, auto-capture iris scanner.[16]

This device is a complete device that captures the near-infrared light-emitting diode (LED) photographs of the eye, stores the data in the form of an algorithm, and then matches it with the stored database. It captures images at an average distance of 5 cm from the image sensor using near-infrared LED illumination. The device is first connected with a USB cable to the device (laptop/mobile phone) compatible with its software. Then, the eye goggle provided by the apparatus is applied to the eye which helps in easy and faster alignment. The eye goggle has been used to block the effect of any room illumination on pupil size. The goggle is just a silicone sleeve that covers the device and touches only the periocular skin but does not have any contact with any ocular structures.

The subject is asked to look at the line circle on the device surface with wide-open eyes, and blue light blinks as soon the sensor detects the iris successfully and stores the processed image after it is autocaptured, completing the whole process in just a few seconds [Fig. 1]. The entire process was carried out in one room, under constant illumination conditions. An HTC Instrument LX-101A Light meter luxmeter was used to keep a check on the room illumination conditions that were kept the same (380–430 lux) during all scanning attempts to avoid bias due to the change in pupil size under different light conditions. The enrollment process was carried out without dilation, and after the enrollment process, the matching was carried out initially with an undilated pupil (that gave our baseline recording). The pupil was dilated by instilling one drop of tropicamide 0.8% and phenylephrine 5% eye drops, and the iris matching process was repeated with the dilated pupil 15 minutes after instilling one drop of mydriatic tropicamide 0.8% and phenylephrine 5% eye drops.

Figure 1.

Image of the eye of a patient (a) before pupil dilation and image of the same eye after pupil dilation (b)

Matching is the process of comparing the newly enrolled data with the baseline enrolled data. Also, the iris is said to be recognized when the machine denotes whether it has matched. The iris scans before and after dilatations were performed in all patients by the same person.

Pupil dilation

Pupil dilation was carried out by the instillation of eye drop tropicamide 0.8% and phenylephrine 5% (w/v), and the iris matching process was repeated with the dilated pupil 15 minutes after instilling the drop.

Clinical evaluation

Clinical evaluation was carried out using slit-lamp ophthalmoscopy for all patients with special emphasis on pupil size and structure. Pupil size was measured using the slit-lamp system, and we did not encounter any change in pupil shape following dilatation. The pupil only was larger in size as compared to the undilated state and maintained its round and regular shape even after dilation.

Statistical analysis

Statistical analysis was performed using SPSS 21 for Windows software. On statistical analysis, values were presented as the mean ± standard deviation for continuous variables or numbers (percentage) for categorical variables. For comparison of the mean of variables (Hamming distance value) within the group, the paired t-test was used. A factorial analysis of variance (ANOVA) test was used to study the effect of pupil size on the Hamming distance value. A P value <0.05 was considered significant.

Results

Demographic feature of the study population

The age of the patients in the study ranged from 12 years to 87 years with a mean age of 56.14 ± 10.91 years. Out of these 321 patients, 173 (53.9%) were males and 148 (46.1%) were females. The mean age of males was 55.84 ± 12.38 years ranging from 12 to 83 years and that of females was 56.45 ± 8.97 years ranging from 24 to 87 years.

Effect of pupil size on the biometric system

All 321 eyes in the study group matched after dilation. The pupil size in the undilated state was less than 4 mm, and in the dilated state, it ranged from 5 mm to 9 mm. Only seven (2.2%) cases had a pupil dilatation of 5 mm. Even though the values were closer to 1 (as in some cases, it was 0.96 as well), because it was less than 1 (threshold value), the status was denoted as matched. The average pupil–iris diameter ratio in the undilated state was 0.33 and that in the dilated state was 0.64. The mean Hamming distance in the undilated state was 0.60 ± 0.07, and it ranged from 0.5 ± 0.84. The mean Hamming distance in the dilated state was 0.79 ± 0.12 with a range from 0.54 to 0.99. The effect of pupil size on the Hamming distance was significant (P value <0.05) [Table 1]. Variable size of pupil dilation was seen over the whole spectrum of eyes under study. No significant difference between the mean Hamming distance was observed between the undilated state and after pupil dilation to 5 mm (P value >0.05). On further scanning of pupils dilated to 6 mm, 7 mm, 8 mm, and 9 mm, a significant change in the mean Hamming distance was observed from the undilated state (P value <0.05), and the pupil size had a significant effect on the Hamming distance. The Hamming distance value grew closer to not matching the threshold cut-off value of 1 as the pupil size progressed [Fig. 2], and the line chart skewed toward the Hamming distance value of 1 after dilation as compared to baseline mean Hamming distance values taken in an undilated state [Fig. 3].

Table 1.

Statistical analysis of the effect of pupil size on Hamming distance

Pupil Size (in mm)	Number of cases (%)	Mean Hamming distance before dilation±SD	Mean Hamming distance after dilation±SD	P
5.00	7 (2.2%)	0.62±0.09	0.61±0.07	0.8
6.00	64 (19.9%)	0.60±0.07	0.67±0.08	<0.001
7.00	113 (35.2%)	0.61±0.08	0.77±0.07	<0.001
8.00	114 (35.5%)	0.60±0.07	0.87±0.07	<0.001
9.00	23 (7.2%)	0.60±0.07	0.96±0.06	<0.001
Total Cases	321 (100%)	0.60±0.07	0.79±0.12	<0.001

Figure 2.

Line chart showing the effect of pupil size after dilation in millimeters on mean Hamming distance values

Figure 3.

Line chart showing the comparison of mean Hamming distance values of the study population before and after pupil dilation

Discussion

Apart from variations in ambient light conditions, changes in pupil size can also occur due to psychological factors,[17,18] diseases,[19] alcohol[20], or drugs.[9,21,22] Also, pupil size is affected by age,[23] with smaller pupils among the elderly population, but there is no direct evidence in the biometric literature thus far, which indicates that changes in recognition performance are induced by changes in pupil size due to aging.[24] The use of iris texture for automated recognition of individuals is commonly known as iris recognition. Changes in pupil size between two iris images of the same person can affect iris recognition performance.

In our study, we observed the effect of pupil dilation on the BIR system. If the Hamming distance value is less than the system’s set threshold value (here, the value of 1), then, it means that the iris has been recognized, which is true for all variable pupil sizes that we encountered after dilation. All eyes in the study matched after dilation, and there were no false matches. It was also seen that as the pupil size increased, the Hamming distance values grew closer to the threshold value of 1 (the threshold for not matching), but even with full dilatation, it was always less than 1 in all eyes. The average pupil–iris diameter ratio in the undilated state was 0.33, and in the dilated state, it was 0.64; this result was consistent with the result of Seyeddain et al.,[15] which was 0.374 and 0.666, respectively, in miosis and mydriasis.

In a prospective, non-comparative cohort study on the effect of cataract surgery and pupil dilation on iris pattern recognition for personal authentication, the authors took images of 15 subjects that were captured before (enrollment) and 5, 10, and 15 min after the instillation of mydriatics before routine cataract surgery. They reported that matching reliability decreased considerably with an increase in pupillary dilation, and pupil dilation may be used to defeat an iris-based authentication system.[10]

A prospective, non-randomized, single-center, cohort study to investigate the reliability of a BIR system for personal authentication after cataract surgery or iatrogenic pupil dilation included 184 eyes that were enrolled in miosis, and the eyes were then evaluated before and after iatrogenic pupil dilation. Out of a total of 184 eyes that were enrolled in miosis, a total of 22 (11.9%) could not be recognized after dilation and therefore needed re-enrollment. They reported that iris recognition was primarily successful in eyes with medically dilated pupils in nearly nine out of ten eyes. No case of false-positive acceptance occurred.[17]

The correct recognition rate in our study was 100%, which was quite high compared to that in previous studies (86.67%[10] and 88.1%[17]). Even though the values were closer to 1 (as in some cases, it was 0.96 as well), because it was less than 1 (threshold value), the status was denoted as matched. No iris texture change was seen after pupil dilation, and in all eyes, the pupil was circular in shape.

The strength of our study was the large sample size of the study population. The notable limitations were that it was a single-centered study and we included only normal healthy eyes of patients who reported for a routine eye check-up. It is recommended that a similar study should be conducted in irregular pupils which develop due to surgical complications or inflammation to find out whether re-enrollment is required for iris recognition-based biometrics in such patients.

Conclusion

In conclusion, we can say that the BIR system is a reliable method for identification and authentication in medically dilated normal pupils.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.