Iris Authentication

Despite the fact that biometric technologies (in particular, the use of the iris to identify a person) are only beginning to gain popularity, the first discoveries in this area were made back in the late thirties of the last century.

• The first that the human eye and its iris can be used for personality recognition, the American eye surgeon, Frank Bursch , thought back in 1936 ^[2] .
• But his idea and development was patented only in 1987. It was not Bursh himself who did this, but ophthalmologists who did not have their own developments — Leonard Flom and Aran Safir ^[2] .
• In 1989, L. Flam and A. Safir decided to turn to John Daugman for help, so that he would develop a theory and recognition algorithms. Subsequently, it is John Daugman who is considered to be the progenitor of this method of biometric authentication ^[2] .
• In 1990, John Daugman first developed a practical method for coding iris structures. The method was patented a little later, in 1993 ^[2] .

• The story of the development of biometric authentication for the iris does not end there. Since 2002, Daugman has released several more articles, each of which more fully reveals and develops this technology. Published articles: Epigenetic randomness, complexity, and singularity of human iris patterns (2001), Gabor wavelets and statistical pattern recognition (2002), The importance of being random: Statistical principles of iris recognition (2003), Probing the uniqueness and randomness of IrisCodes : Results from 200 billion iris pair comparisons (2006), New methods in iris recognition (2007), Information Theory and the IrisCode (2015).

In this case, the iris is considered as a physiological parameter - a round plate with a lens in the center, one of the three components of the choroid (middle) of the eye .

The iris is located between the cornea and the lens and serves as a kind of natural diaphragm that regulates the flow of light into the eye. The iris is pigmented, and it is the amount of pigment that determines the color of a person’s eyes ^[3] .

In its structure, the iris consists of elastic matter - the trabecular network . This is a mesh formation that forms at the end of the eighth month of pregnancy. The trabecular network consists of grooves, comb screeds, furrows, rings, wrinkles, freckles, blood vessels and other features. Due to such a number of components, the “pattern” of the network is rather random, which leads to a high probability of uniqueness of the iris. Even for twins, this parameter does not completely coincide ^[4] .

Despite the fact that the iris of the eye can change its color up to one and a half years from the moment of birth, the pattern of the trabecular network remains unchanged throughout a person’s life. An exception is serious injury and surgery ^[4] .

Due to its location, the iris is a fairly protected part of the organ of vision, which makes it an excellent biometric parameter.

Most of the iris identification systems and technologies currently in operation are based on the principles proposed by J. Daugman in the article “High confidence visual recognition of persons by a test of statistical independence” ^[5] .

The process of personality recognition using the iris can be divided into three main stages: digital image acquisition, segmentation and parameterization. Each of these steps will be discussed in more detail below.

Image Acquisition

The authentication process begins with obtaining a detailed image of the human eye. They try to make an image for further analysis in high quality, but this is not necessary. The iris is so unique that even a fuzzy picture will give a reliable result. For this purpose, use a monochrome CCD camera with dim lighting, which is sensitive to infrared radiation. Usually a series of several photos is taken due to the fact that the pupil is sensitive to light and constantly changes its size. The backlight is unobtrusive, and a series of pictures is taken in just a few seconds. Then, one or several are selected from the obtained photographs and segmentation is started ^[6] .

Segmentation

Segmentation deals with the separation of the image of the outer part of the eye into separate sections (segments). In the process of segmentation, the iris is first found in the photograph obtained, the inner border (near the pupil ) and the outer border (border with the sclera ) are determined. After that, find the boundaries of the upper and lower eyelids, and also exclude the accidental application of eyelashes or glare (from glasses, for example) ^[7] .

The accuracy with which the borders of the iris are determined, even if they are partially hidden for centuries, is very important. Any inaccuracy in the detection, modeling and further presentation of the iris can lead to further failures and inconsistencies ^[7] .

After determining the boundaries, the image of the iris must be normalized. This is not a very obvious, but necessary step, designed to compensate for changes in the size of the pupil. In special cases, normalization is a transition to the polar coordinate system . John Daugman applied and described this in his early works ^[5] . After normalization using pseudo-polar coordinates, the selected image area becomes a rectangle, and the radius and center of the iris are estimated ^[8] .

Parameterization

During parameterization of the iris, a control region is extracted from the normalized image. Two-dimensional Gabor waves are applied to each point of the selected region (other filters can be applied, but the principle remains the same) in order to extract phase information. The undoubted advantage of the phase component is that, unlike the amplitude information, it does not depend on the contrast of the image and lighting ^[9] .

The resulting phase is usually quantized with 2 bits, but another quantity can be used. The total length of the description of the iris, thus, depends on the number of points at which the phase information is found, and the number of bits required for encoding. As a result, we get the iris pattern, which will be bit by bit checked with other patterns in the authentication process. A measure by which the degree of difference between the two irises is determined is the Hamming distance ^[9] .

Some countries have already begun to develop a program, part of which will be biometric authentication based on the iris. It is planned that with the help of this innovation the problem of fake passports and other identity cards will be solved. The second goal is the automation of passport control and customs screening at the entrance to the country using biometric passports ^[10] .

Since 2004, a no less complicated project has been operating in the UK - IRIS (Iris Recognition Immigration System). Under this program, about a million tourists from abroad, often traveling to the UK, may not have to provide their documents at airports for identification. Instead, a special video camera checked their iris with an already formed base. In 2013, this project was abandoned in favor of biometric passports, where information about the iris of the eye is also recorded ^[10] .

In order for this or that characteristic of a person to be recognized as a biometric parameter, it must meet five specially developed criteria : universality, uniqueness, constancy, measurability and acceptability.

The universality of the iris is not in doubt. Also, from clinical studies revealed its uniqueness and stability ^[11] . With regard to measurability, this point is confirmed by the mere existence of articles and publications by J. Daugman ^{[5] [12] [13]} . The last point, the question of acceptability, will always be open, as it depends on the opinion of the community.

A comparison table of biometric authentication methods, where H - High, M - Medium, L - Low ^[14] :

At the moment, biometric technology has not yet been created that would fully comply with all five points. But the iris is one of the few parameters that correspond to most ^[15] .

Method Accuracy

In biometrics, when calculating the accuracy of a method, errors of the first and second kind (FAR and FRR) are taken into account ^[16] .

FAR (False Acceptance Rate) - the probability of false tolerance of the object.

FRR (False Rejection Rate ) - the probability of a false rejection of an object.

These two concepts are closely related, since a decrease in one error leads to an increase in the other. Therefore, developers of biometric systems are trying to come to some kind of balance between FAR and FRR ^[17] .

One of the methods for determining the accuracy of a system that involves errors of the first and second kind is the method of constructing an ROC curve .

The ROC curve is a graphical representation of the relationship between the FAR and FRR characteristics with varying the threshold of sensitivity (threshhold) ^[18] . The sensitivity threshold determines how close the current sample should be to the template in order to consider them matching. Thus, if a small threshold is selected, the number of false tolerances increases, but the probability of a false deflection of the object decreases. Accordingly, when choosing a high threshold, everything happens the other way around ^[17] .

Sometimes they introduce a new parameter - EER.

EER (Equal Error Rate) is a value that characterizes the error level of the biometric method, at which the FAR and FRR values ​​are equal. The smaller this parameter, the more accurate the system. The ERR value is recognized using the ROC curve described above ^[19] .

With regard to accuracy, directly, authentication by iris, a good source is the book "Handbook of Iris Recognition" . This paper describes an experiment in which several types of biometric technologies were compared. Based on these studies, the accuracy of iris authentication reaches 90% ^[20] .

In another work, it was found that the FAR value of this method under certain conditions can take values ​​from 1% or lower, and the FRR value is constant and tends to zero (0.00001%) ^[21] .

In turn, the FAR and FRR values ​​directly depend on the processes of obtaining and processing the iris image. A large role in this is played by the filters used in the segmentation process. From the table below, you can see how changing one filter affects the final result ^[22] .

The table of parameters FAR (%), FRR (%) and EER (%) depending on the choice of filter ^[22] :

Retina Authentication Comparison

Most often, people confuse physiological parameters such as the retina and the iris. More often, they combine two concepts into one. This is a huge misconception, since the retinal authentication method involves studying the fundus. Due to the length of this process and the large installation size, this type of authentication can hardly be called public and convenient. In this, retinal biometric authentication loses to iris authentication ^[23] .

• L. Flom, A. Safir US Patent 4641349
• R.M. Ball, J.H. Connell, S. Pankanti, N.K. Ratha, E.U. Senor. Guide to biometrics. - M .: Technosphere, 2007 .-- S. 20 - 63. - 368 p. - ISBN 978-5-94836-109-3 .
• Khalid Saeed, Tomomasa Nagashima. Chapter 3. Iris Pattern Recognition with a New Mathematical Model to Its Rotation Detection // Biometrics and Kansei Engineering. - Springer Science & Business Media, 2012 .-- P. 43 - 65. - 276 p. - ISBN 978-1-461-45607-0 .
• Anil Jain, Arun A. Ross, Karthik Nandakumar. Chapter 4 Iris Recognition // Introduction to Biometrics .. - Springer Science & Business Media, 2011 .-- P. 141-175. - 276 p. - ISBN 978-0-387-77326-1 .
• Rajesh M. Bodade, Sanjay Talbar. Introduction to Iris Recognition // Iris Analysis for Biometric Recognition Systems. - Springer, 2014 .-- P. 3 - 5. - 109 p. - ISBN 978-8-132-21853-1 .
• Anil Jain, Ruud Bolle, Sharath Pankanti. Recognizing Persons by Their Iris Patterns // Biometrics: Personal Identification in Networked Society. - Springer Science & Business Media, 2006. - P. 102 - 122. - 411 p.
• José Ruiz-Shulcloper, Walter Kropatsch. An Alternative Image Representation Model for Iris Recognition // Progress in Pattern Recognition, Image Analysis and Applications. - Springer Science & Business Media, 2008. - P. 86 - 93. - 814 p.
• AJ Mansfield, JL Wayman. Definitions // Best Practices in Testing and Reporting Performance of Biometric Devices: Version 2.01. - Center for Mathematics and Scientific Computing, National Physical Laboratory, 2002. - P. 7 - 8. - 32 p.
• Mark J. Burge, Kevin Bowyer. Fusion of Face and Iris Biometrics // Handbook of Iris Recognition. - Springer-Verlag London, 2013 .-- P. 234. - 399 p.
• J. Daugman. High confidence visual recognition of persons by a test of statistical independence // IEEE Transactions on Pattern Analysis and Machine Intelligence. - 1993. - Vol. 15 , no. 11 . - P. 1148 - 1161 .
• J. Daugman. How iris recognition works // IEEE Transactionson Circuits and Systems for Video Technology. - 2004. - Vol. 14 , no. 1 . - P. 21 - 30 .
• J. Daugman. New Methods in Iris Recognition // IEEE Trans. Systems, Man, and Cybernetics. - 2007. - Vol. 37 , no. 5 . - P. 1167 - 1175 .
• J. Daugman. Probing the Uniqueness and Randomness of IrisCodes: Results From 200 Billion Iris Pair Comparisons (English) // IEEE Transactionson Circuits and Systems for Video Technology. - 2007, january. - Vol. 94 , no. 11 . - P. 1927 - 1935 .
• Anil Jain, Arun Ross and Salil Prabhakar. An Introduction to Biometric Recognition // IEEE Transactions on Circuits and Systems for Video Technology. - 2004. - Vol. 14 , no. 1 . - P. 4 - 20 .
• Dr. Chander Kant, Sachin Gupta. Iris Recognition: The Safest Biometric (Eng.) // An International Journal of Engineering Sciences ISSN. - 2011. - Vol. 4 . - P. 265 - 273 .
• Alekseev V.N., Astakhov Yu.S., Basinsky S.N. Chapter 2. Anatomy of the organ of vision // Ophthalmology: Textbook for students. honey. universities / E.A. Egorov. - M .: GEOTAR-Media, 2008 .-- S. 12 - 29. - 240 p.
• Pavelieva E.A., Krylov A.S. Algorithm for comparing images of the iris of the eye based on key points (Russian) // Informatics and its applications. - 2011. - T. 5 , No. 1 . - S. 68 - 72 .