International Conference on Current Trends in Computer, Electrical, Electronics and Communication (ICCTCEEC-2017)

An Approach for Secure Identification and

Authentication for Biometrics using Iris

Chandrashekar M Patil Sushmitha Gowda

Prof, Dept. of ECE Department Of ECE
VVCE, Mysore VVCE, Mysore
[email protected] [email protected]

Abstract— One of the most accurate biometric authentication benefits of being contactless, low mis-identification rate and
methods is iris pattern recognition. Basically iris recognition is a no preceding information from user’s side is needed to start
method of biometric recognition and authentication that uses the usage of this method.
recognition strategies on images of a people eyes. This paper
presents a secure authentication using iris recognition system. A Iris is well blanketed from the encircling environment
usual recognition system has distinctive phases like enrollment because its miles an inner a part of the eye, and its patterns are
phase and authentication phase. In this proposed work CASIA strong during the lifestyle. The iris is a skinny circular
V4 iris is considered as a database. The flow of work starts with anatomical form in the attention and it is an annular part
pre-processing, feature extraction and classification. This among black pupil and the white sclera, the iris contain
procedure is implemented using evaluation of various iris pattern variable sized hole referred to as pupil. It presents visible
recognition features by using Histogram of Oriented Gradients function denoted as the texture of the iris. The iris’s
(HOG), Gray Level Co-occurrence Matrix (GLCM), Hausdroff characteristic is to manipulate the diameter and size of the
Dimensions (HD), Biometric Graph Matching (BGM) and 2D - pupil and therefore it controls the quantity of light that
Gabor filter techniques. Then, a State Vector Machine (SVM) , progresses to the retina.
Weighted Euclidian Distance (WED) and Jaccard coefficient and
Dice coefficient classifier is used to verify and authenticate the
patterns. Accordingly, SVM is regarded as most distinctive and
most efficient technique for the proposed iris identification and
authentication system.

Keywords—Iris authentication; secure identification; classifier;

enrollment; biometric recognition.

I. INTRODUCTION
Fig. 1. Front veiw of iris
In today’s statistics technology, security for systems is
turning into increasingly more essential. The range of The front view of the iris is shown in Figure 1. To
structures which have been compromised is ever increasing manipulate the amount of mild getting into the eye, the
and authentication plays a primary role as a first line of muscular tissues associated with the iris (sphincter and dilator)
defence against intruders. Modern day e-safety are in critical either increase or agreement the centre aperture of the iris
need of locating accurate, comfortable and cost-effective referred as the pupil.
alternatives to passwords and personal identity numbers as This paper uses three different hybrid combination
financial losses increase dramatically year over year from pc- techniques for Feature extraction & Classification of iris
based fraud including computer hacking and identity theft. recognition module as a user security are utilized to design
Biometric answers deal with those fundamental troubles, proposed system.
because an individual's biometric statistics particular and
cannot be transferred.
II. LITERATURE REVIEW
Biometrics refers to the identification or authentication of It has been known that eye is a abundant source of
an man or woman primarily based on positive particular information, it has many distinguish features, which can be
capabilities or traits which has capability to differentiate utilized for user identification.
between legal person and an imposter. The two classes of
biometric identifiers mainly physiological and behavioral Himanshu rai, Anamika yadav [5] uses a circular hough
traits. Iris, Fingerprint, DNA, and many others belong to the transformation for segmentation of iris images. Parabolic
former kind of biometric identifiers at the same time as typing hough transformation and trimmed median filter is used for
rhythm, gait, voice, and so on. belong to the latter. Out of all eyelid detection. While, Haar wavelet and 1d log Gabor filter
the numerous physical traits to be had, irises are one of the extract the features and for recognition of iris patterns it uses a
more accurate physiological traits that may be used. It has the aggregate of support vector device and hamming distance. The

978-1-5386-3243-7/17/$31.00 ©2017 IEEE

421
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on January 16,2023 at 09:32:34 UTC from IEEE Xplore. Restrictions apply.
 International Conference on Current Trends in Computer, Electrical, Electronics and Communication (ICCTCEEC-2017)

drawback in this method is requirement of calculation is very applied on images as a results it increases contrast of an
big, huge quantity of storage and also excessive price is original image.
necessary for computation.
Dr H.B Kekre et.al [13] proposed an iris recognition
device that's based totally on vector quantization which
incorporates Kekre’s rapid codebook generation algorithm. Its
overall performance is in comparison with the discrete cosine
transform (DCT) as proposed VQ based system does now not
want any pre-processing and segmentation of the iris. The
KFCG algorithm requires 99.99% less computations and gives
accuracy of 89.10% outperforming DCT accuracy around
66.10%.
Ya-ping Huang et.al [8] proposed a iris recognition
technique which constructs primary features for training set by
way of impartial factor analysis, which determines the centre
of each magnificence with the aid of competitive getting to
know mechanism and in the end acknowledges the pattern
primarily based on euclidean distance. No restriction for photo Fig. 2. The Proposed Algorithm
capture due to representation of length and rotation invariance. From the eye photo the iris element is separated by the use
But, the major drawback of this technique is the algorithm of image segmentation approach. The edges of the iris internal
makes use of all patterns of each class as a whole to estimate and external boundaries are located and detected by the use of
ICA basic functions and while a brand new class is added all canny edge detector. Hough transform is used for locating out
of the patterns should gain knowledge of again. the circles inside the iris picture and decreases a pixel at the
boundary of the circle. The location of the edge can be
III. METHODOLOGY received despite of the nonappearance of few pixels.
Figure 2 is the proposed algorithm for iris recognition Normalization is applied for unwrapping the annular common
system. This algorithm is implemented in two phases : iris areas into regular and constant rectangular form.

™ Enrollment Phase (Training phase) : Proposed feature extraction and classification technique is
are in table 1.
x Capture the image of an every person to be TABLE I : Feature Extraction & Classification Techniques
authenticated.
Proposed Feature extraction Classification /
x Apply pre-processing methods to improve Methodology techniques Matching
the quality of images. algorithm
01 GLCM, HOG, DCT (WED)
x Extract various features from iris images
using different hybrid combinations 02 GLCM, HD, BGM (SVM)
technique and store the extracted feature 03 2D - Gabor filter Jaccard and
vector in a database. Dice coefficient
™ Authentication Phase (Testing phase) :
A. Feature extraction
x Capture the entity/person image who need to
get access to the system. 1. Gray Level Co-occurrence Matrix (GLCM)
x Apply all previous step on the captured test A Gray Level Co-occurrence Matrix (GLCM) or co-
image including pre-processing until getting occurrence is a matrix or distribution that is described over a
the all required feature vector of an test photo to be the distribution of co-taking place values at a
image. given offset. In the beginning the co-occurrence matrix is built
based on route and distance among photo pixels. Then features
x Match the image feature vector of the test are extracted from the matrix as the texture representation.
image with the template stored on the From the co-occurrence matrix various functions may be
database by using different classifiers. extracted like Energy, Contrast, Entropy, Correlation
coefficient, Homogeneity and many other. GLCM approach
x If match, identity is verifies and turn out to be efficiently used in lots of felids of feature
authenticated else rejected. extraction. Twenty one GLCM functions were extracted from
At last, all the three methods are compared using every block and mixed to shape a feature vector of 1x88
performance metrics (Accuracy and Error rate). values for every segmented iris picture.
The captured image is enrolled with its owner identity in
the enrollment phase. Histogram equalization technique is

422
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on January 16,2023 at 09:32:34 UTC from IEEE Xplore. Restrictions apply.
 International Conference on Current Trends in Computer, Electrical, Electronics and Communication (ICCTCEEC-2017)

2. Histogram of Oriented Gradients (HOG) maximizing the space of both classes from the hyper-plane
This feature selection counts the occurrences of gradient using the Gaussian radial foundation kernel. The margin
orientation in localized quantities of an picture and returns denotes the maximum width inside the hyper planes.
extracted HOG features. Returned features encode 3. Jaccard Coefficient and Dice Coefficient
neighborhood form data from regions inside an photograph.
It's far used for the motive of category, detection and tracking. Distance based similarity measure is primarily taken into
consideration as a critical parameter for the classification and
3. Discrete Cosine Transforms (DCT) measuring the similarity or dissimilarity between two or extra
A Discrete Cosine Transform (DCT) is an extensively used vectors/objects.
set of rules to transform statistics into sets of frequencies and
it is a real transform which has fantastic advantage in energy IV. EXPERIMENTAL RESULTS
compaction. Most of the signal energies in a picture are
Many experiments have been conducted to check the
concentrated at decrease (lowest) frequencies, so there's only
performance of the proposed algorithm which was
requirement of simplest decrease (low) frequency additives for
implemented using MATLAB and Graphical User Interface
recognition.
(GUI).
4. Hausdroff Dimensions (HD)
The proposed algorithm was tested on CASIA- Iris- Syn
In order to extract iris features the Hausdorff Dimension is Version 4 dataset. Firstly, an original input mage shown in
a completely efficient approach. This method measures degree figure 3.a. after histogram equalization, iris and pupil
of roughness or texture of the image. From the gray-degree of localization the result is shown in figure 3.b. and 3.c.
the iris photograph binary photograph in obtained. Edge GLCM, DCT, HOG, HD, BGM, 2D-gabor filter are used
detection approach is hired to hint out the boundaries of the to extract features from each preprocessed images. Magnitude
pixels. of Gabor filter of an input image from dataset is as shown in
the Hausdorff Distance is defined as follows : figure 4.
H (M,N) = max (h( M , N ) , h( N , M ))
5. Biometric graph matching (BGM)
BGM algorithm makes use of graph topology to outline
distinct characteristic values of the iris templates. An iris
matching machine has been presented based on the biometric
graph matching set of rules. The iris picture was segmented. A (a) (b)
set of photos are taken and locating out the functions values.
6. 2D - Gabor Filter
The Gabor filters are broadly used in the characteristic
extraction of the photograph. The Gabor filter is the form of
sine wave modulated by the way of Gaussian coefficient. The
Gabor filters are useful for extracting the local shape and
global statistics' of the iris. This filters are generally tunable (c)
band pass, multi scale and multi decision filter. Fig. 3. Iris recognition results. (a).Input Image,( b).Histogram
equalization, (c).Iris and pupil localization
B. Classification
1. Weighted Euclidean Distance (WED)
The Euclidean distance is the basis of many measures of
similarity and dissimilarity. The weighted Euclidean distance
(WED) is used generally to evaluate unique templates
particularly if the template is composed of integer values.
WED offers a degree of how similar a group of values among
the two templates.
2. State Vector Machine (SVM)
SVMs are in general two-magnificence classifiers that
have been proven to be an attractive and extra systematic Fig. 4. Magnitude of Gabor filter of an input image
method to study linear or non-linear choice boundaries and its
is used for classification of data points. There are two types in In order to classify and authenticate the person in the
the data points which can be real or imposter. The effective proposed algorithm three different classifiers have used such
method to classify the two data points is hyper planes. as SVM, WED and Jaccard and Dice co-efficient. This has
Consequently classifier constructs an optimal isolating hyper- been implemented in GUI as shown in figure 5.
plane between the classes within the dataset by way of

423
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on January 16,2023 at 09:32:34 UTC from IEEE Xplore. Restrictions apply.
 International Conference on Current Trends in Computer, Electrical, Electronics and Communication (ICCTCEEC-2017)

V. CONCLUSION AND FUTHER WORK

This proposed work gives a simple and stable solution for
iris recognition system for secure authentication and algorithm
focus on the methods for rapid and accurate iris identification
and authentication. The various features extraction of an iris
images are done by using hybrid combinations of gray level
co-occurrence matrix (GLCM), Histogram Of Oriented
Gradients (HOG), Hausdroff Dimensions (HD), Biometric
Graph Matching (BGM), 2D Gabor filter methods and
classifiers that are SVM, WED , Jaccard and Dice coefficient
gives effective results with less number of false detections.
Future work can aim to enhance the system to cope with
image degradation by noise, duty and glasses. Coloration
Fig. 5. Recognition result function can be also used to increase the recognition accuracy.
There is scope to work on developing an algorithm based on
Table II shows execution time of three different techniques multimodal biometrics devices which will integrate both the
and performance metric is as shown in figure 6. All three biometric traits derived from one or modalities such as palm
methods give satisfying result for recognition process but the print and iris which provide excessive level of security.
performance metrics, accuracy and error rate differ among
them. The SVM gives the best accuracy and error rate of 95 %
References
and 5 % respectively than all three methods because WED and
JDC method uses near distance method to match the template. [1] Gagan R and Lalith S, Elliptical Sector Based DCT Feature Extraction
for IRIS Recognition,2015.
Other two methods follows behind such as WED and Jaccard
[2] Chun-Wei Tan , Ajay Kumar, Accurate Iris Recognition at a distance
and dice co-efficient methods produces 90 % accuracy and 7 using stabilized Iris encoding and Zernike Moments Phase Features,
% error rate. 2014.
TABLE II: Elapsed time for Feature Extraction & Classification [3] Chung-Chih Tsai, Heng-Yi Lin, Jinshiuh Taur ,Chin-Wang Tao, Iris
Recognition Using Possibilistic Fuzzy Matching on Local
Features,2012.
Methodology Time taken for feature Time taken for [4] Shengnan Sun, Shicai Yang, Lindu Zhao, Noncooperative Bovine Iris
extraction Recognition via SIFT”, 2013.
classification
01(WED) 1 minute 25 seconds 3 seconds [5] Himanshu Rai , Anamika Yadav, Iris recognition using combined
support vector machine and Hamming distance approach, Expert
Systems with Applications Volume 41, Issue 2, 2014.
02(SVM) 1 minute 1 second [6] W. W. Boles and B. Boashash , A Human Identification Technique
Using Images of the Iris and Wavelet Transform 1998.
03(JDC) 48 seconds 1 second [7] J .Daugman, High Confidence Visual Recognition of Persons by a Test
of Statistical Independence, 1993.
[8] Ya-Ping Huang, Si-Wel Luo and En-Yi Chen, An Efficient Iris
Recognition System, 2002.
[9] C. Chin, A. Jin, and D. Ling, High Security Iris Verification System
based on Random Secret Integration,”Proceedings of International
conference on Computer Vision and Image Understanding, 2005.
[10] M Vatsa, R Singh, and A Noore ,Reducing the False Rejection Rate of
Iris Recognition Using Textural and Topological Features, 2008.
[11] G. Savithiri, A.Murugan, Performance Analysis on Half Iris Feature
Extraction using GW, LBP and HOG, 2011.
[12] Shirke Swati D., Shirke Suvarna D. And Gupta Deepak, Iris
Recognition, 2012 .
[13] H. B. Kekre, Tanuja K. Sarode, Vinayak Ashok Bharadi, Abhishek A.
Agrawal, Rohan J. Arora, and Mahesh C. Nair, Iris Recognition Using
Discrete Cosine Transform and Kekre’s Fast Codebook Generation
Algorithm.

Fig. 5. Performance Metrics

424
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on January 16,2023 at 09:32:34 UTC from IEEE Xplore. Restrictions apply.