76 IJCSNS International Journal of Computer Science and Network Security, VOL.16 No.

9, September 2016

A Novel Retina based Biometric Privacy using Visual

Cryptography
M. Suganya1 and K. Krishnakumari2
1
Research Scholar, Department of Computer Science, Rathnavel Subramaniam College of Arts & Science, Sulur, India.
2
Director, Department of MCA, Rathnavel Subramaniam College of Arts & Science, Sulur, India.

ABSTRACT are available such as facial thermo gram, hand vein, odor,
Biometrics deal with automated methods of identifying a person ear, hand geometry, fingerprint, face, retina, iris, palm
or verifying the identity of a person based on physiological or print, voice and signature. Among those retina recognition
behavioral characteristics. Preserving the biometric privacy such is one of the most promising approach because of stability,
as digital biometric data (eg. face, iris, retina and fingerprint) is uniqueness and noninvasiveness.
very important nowadays. The retinal biometrics is considered
one of the most accurate and robust methods of the identity
Biometrics systems are more consistent and more user
verification. The unique retina features of an individual can be friendly. Still there are certain issues particularly the
presented in a compact binary form which can be easily security facet of both biometric system and biometric data.
compared with the reference template to confirm identity. As As template is stored in centralized database, they are
biometric template are stored in the centralized database, due to vulnerable to eavesdropping and attacks. Thus alternative
security threats biometric template may be modified by attacker. protection mechanisms need to be considered. For these
If biometric template is altered authorized user will not be reasons various researches have been made to protect the
allowed to access the resource. To deal this issue visual biometric data and template in the system by using
cryptography schemes can be applied to secure the iris template. cryptography, stenography and watermarking. In this work
Visual cryptography is a secret sharing scheme where a secret
image is encrypted into the shares which independently disclose
a system is proposed by applying visual cryptography
no information about the original secret image. The combination technique to biometric template (retina). Visual
of biometrics and visual cryptography is a promising information cryptography technique has been applied on to the retinal
security technique which offers an efficient way to protect the template to make it secure from attack in centralized
biometric template. Visual cryptography provides great means database as well as extra layer of authentication to the
for helping such security needs as well as extra layer of users.
authentication.
Keywords: 1.1 Modules in Biometric System
Biometric, Visual cryptography, Retina authentication,
Enrollment, Authentication There are basically two phases in biometric system. There
are enrollment phase and authentication phase. In these
two phases there are four modules. The sensor module is
1. Introduction used in extracting the biometric data which may be image,
audio or video. The feature extraction module is used in
Security of data has been a major issue from many years. obtaining the template that is generated from the features
Using the age old technique of encryption and decryption of the biometric data. Each feature is labeled with a user’s
has been easy to track for people around. Providing
identity. The Matching module is used in authentication
security to data using new technique is the need of the
phase, where the template data is compared with data
hour. This project uses the technique of Visual which is obtained from user and that it estimates the
cryptography and providing biometric authentication. For similarity between these data. These similar elements are
automated personal identification biometric authentication processed in Decision making module which is used to
is getting more attention. Biometrics is the detailed identify the individual.
measurement of human body. Biometrics deal with
This work is organized as follow: Related work for
automated methods of identifying a person or verifying security enhancement of biometrics system and various
the identity of person based on physiological or behavioral visual cryptography schemes are discussed in section2,
characteristics. There are various applications where section 3 presents the proposed system, experiments and
personal identification is required such as passport. results are shown in section 4, and section 5 concludes the
Controls, computer login control, secure electronic work.
banking, bank ATM, credit cards, airport, mobile phones,
health and social services, etc. Many biometric techniques

Manuscript received September 5, 2016

Manuscript revised September 20, 2016
IJCSNS International Journal of Computer Science and Network Security, VOL.16 No.9, September 2016 77

2. Literature Survey security, tampering attacks can be mounted. The results of

a corresponding attack against a multimodal iris
Ross and Asem (2011) explores the possibility of using recognition scheme show, that in this environment either
visual cryptography for imparting privacy to biometric semi-fragile watermarking or additional classical
data such as fingerprint images, iris codes, and face cryptographic means need to be applied to secure the
images. In the case of faces, a private face image is system against the demonstrated attack which is presented
dithered into two host face images (known as sheets) that by Hammerle et al (2011) Bringer et al (2011) introduces
are stored in two separate database servers such that the a new method to identify someone using his biometrics in
private image can be revealed only when both sheets are an encrypted way. Our construction combines Bloom
simultaneously available; at the same time, the individual Filters with Storage and Locality-Sensitive Hashing. We
sheet images do not reveal the identity of the private apply this error-tolerant scheme, in a Hamming space, to
image. The purpose of Rajanwar et al (2014) is to protect achieve biometric identification in an efficient way.
biometrics data from the various attacks. We are using the Arkala et al (2011) represent the retina vessel pattern as a
concept of visual cryptography, where cryptography is the spatial relational graph, and match features using error-
concept of sending and receiving encrypted messages and correcting graph matching. They study the distinctiveness
that can be decrypted by the authorized persons with the of the nodes (branching and crossing points) compared
required keys only. Labati et al (2012) focuses on the most with that of the edges and other substructures (nodes of
important privacy issues related to the use of biometrics, it degree k, paths of length k).
presents actual guidelines for the implementation of Monwar et al (2011) utilize the physiological attributes
privacy-protective biometric systems, and proposes a (face, ear and iris) along with soft biometric information
discussion of the methods for the protection of biometric (gender, ethnicity and eye color). A fuzzy fusion
data. A comprehensive survey of biometric cryptosystems mechanism for robust and reliable multimodal biometric
and cancelable biometrics is presented by Rathgeb and based security systems is developed. Barkhoda et al
Andreas (2011). State-of-the-art approaches are reviewed (2011) proposed a novel human identification method
based on which an in-depth discussion and an outlook to based on retinal images. The proposed system composed
future prospects are given. In this correspondence, of two main parts, feature extraction component and
Simoens et al (2012) analyze the vulnerabilities of decision-making component. Instead of optimally aligning
biometric authentication protocols with respect to user and two iris-codes by maximizing the comparison score for
data privacy. The goal of an adversary in such context is several bit shifts utilizes the total series of comparison
not to bypass the authentication but to learn information scores, avoiding any information loss. Blanton and Paolo
either on biometric data or on users that are in the system. (2011) develop and implement the first privacy-preserving
The design of single-use biometric security systems is identification protocol for iris codes. We also design and
analyzed by Lai et al (2011) from an information theoretic implement a secure protocol for fingerprint identification
perspective. A fundamental trade-off between privacy, based on FingerCodes with a substantial improvement in
measured by the normalized equivocation rate of the the performance compared to existing solutions. Qamber
biometric measurements, and security, measured by the et al (2012) present a system for recognition based on
rate of the key generated from the biometric vascular pattern of human retina. The proposed algorithm
measurements, is identified. consists of three stages; i.e. preprocessing, feature
Hao et al (2013) propose a method using color histogram extraction and finally the matching process.
as the trait of retina biometric. The color histogram has
shown the feature of deformation invariant. The
enhancement using discrete wavelet transform 3. Research Methodology
decomposition minimizes the difference within the
individuals and increases the correlation between the This section gives the brief explanation of visual
retinal images of the same person. Lajevardi et al (2013) cryptography and methodology used for retina biometric
presents an automatic retina verification framework based privacy.
on the biometric graph matching (BGM) algorithm. The
retinal vasculature is extracted using a family of matched 3.1 Visual Cryptography
filters in the frequency domain and morphological
operators. Then, retinal templates are defined as formal Cryptography is the art of sending and receiving encrypted
spatial graphs derived from the retinal vasculature. Several messages that can be decrypted only by the sender or the
receiver. Encryption and decryption are accomplished by
multimodal biometric schemes have been suggested in
using mathematical algorithms in such a way that no one
literature which employs robust watermarking in order to
embed biometric template data into biometric sample data. but the intended recipient can decrypt and read the
In case robust embedding is used as the sole means of message. Naor and Shamir introduced the visual
cryptography scheme (VCS) as a simple and secure way
78 IJCSNS International Journal of Computer Science and Network Security, VOL.16 No.9, September 2016

to allow the secret sharing of images without any A. Enrollment

cryptographic computations. This scheme is referred to as
the k-out-of-n VCS which is denoted as (k,n)VCS. Given The administrator will collect the eye image of the eligible
an original binary image, it is encrypted in n images, such users those are having access to secure resource. The
that enrolled eye image is required to be processed so
𝑇𝑇 = 𝑆𝑆ℎ1 ⊕ 𝑆𝑆ℎ2 ⊕ 𝑆𝑆ℎ3 ⊕ … … … … … 𝑆𝑆ℎ𝑛𝑛 characteristic retinal features can be extracted. Three steps
(1) that are: segmentation, normalization, and feature
extraction are performed as conferred below:
• Segmentation is performed to extract the retinal
Where Å is a Boolean operation, Shi, hi 𝐼𝐼̂ 1, 2,….k is an image from the eye.
image which appears as white noise, k d” n, and n is the
• Normalization of retina region is carried out
number of noisy images. It is difficult to decipher the
using sheet model. This model remaps each pixel
secret image T using individuals Shi‘s. The encryption is
within the retina region to a pair of polar
undertaken in such a way that k or more out of the n
coordinates.
generated images are necessary for reconstructing the
original image T. In the case of (2, 2) VCS, each pixel P in • Feature extraction is done by convolving the
the original image is encrypted into two sub pixels called normalized retina pattern into one dimensional
shares. For biometric privacy, here 2-out-of-2 scheme is wavelet.
using.

Fig. 2 User Enrollment

Fig I. Encoding A Binary Pixel P Into 2 Shares A And B

In this scheme for sharing a single pixel p, in a binary

image Z into two shares A and B is illustrated in Table I.
If p is white, one of the first two rows of Table 1 is chosen
randomly to encode A and B. If p is black, one of the last
two rows in Table 1 is chosen randomly to encode A and Fig. 3 User Authentication
B. Thus, neither A nor B exposes any clue about the
binary color of p. When these two shares are In the existing system generated template is stored in the
superimposed together, two black sub-pixels appear if p is database. As Nalini K. Ratha et al pointed out that the
black, while one black sub-pixel and one white sub-pixel stored template in the database attacker may try to alter
appear if p is white as indicated in the rightmost column in result in authorization for unauthorized users, or denial of
Table 1. Based upon the contrast between two kinds of service for the authenticated user related with the
reconstructed pixels can tell whether p is black or white. corrupted template. Here retina template is protected by
applying visual cryptography. For securing retina feature
3.2 Proposed Approach template, the template and another secret binary image
which is chosen by system administrator is given as input
As protecting template in the database securely is one of to the visual cryptography. Two random shares are created
the challenges in any biometric system. Here visual with the help visual cryptography scheme. For sharing two
cryptography technique is applied to retina authentication secret images R1 is retina template image (generated from
system. In this system there are two modules: Enrollment feature template) and R2 is secret image, two shares S1
module and Authentication module. For accessing any and S2 are generated.
secure resource by authenticated users this system can be One share is stored in the database along with user login
used. and other given to user on ID card along with login.
Enrollment process is shown in the figure 2. As the visual
IJCSNS International Journal of Computer Science and Network Security, VOL.16 No.9, September 2016 79

cryptography techniques guarantee that no information is generated. Now this feature template is matched with
revealed by one share alone, this provides security to the retina feature of newly provided retina image.
retina template in the database.

B. Authentication

For authentication user will provide share in the form of

ID card. System finds the corresponding share from
database. By stacking two shares first R1 retina template
image is created. And from this image retina feature (a) (b) (c)
template is generated. The new eye image supplied by
user will be processed with three steps: segmentation, Fig. 4 (a) Retina Image (b) Retina segmentation (c) Extracted feature
normalization and feature extraction which generates template
retina feature template. Then these two feature templates
are matched. If features match access is granted else the
verification fails. Authentication process is shown in
figure 3.

4. Experimental Results
The main intent of this work is providing security to the Fig. 5 (a) Retina template image (b) Share1 (c) Share2 (d) Result of
retina template in the database. To build this system, superimposing of share1 and share 2
MATLAB platform is selected because of powerful inbuilt
mathematical, signal and image processing functions of In this experiment, the possibility of exposing the identity
visual cryptography. Retina images are taken from DRIVE of the secret image by using the sheet images in the
standard database. matching process is investigated. For this experiment, the
Since the proposed technique was devised for binary eye sheet images for three different Retina samples of the
images, a threshold value was used to generate the binary same subject were first computed. Next, the reconstructed
image for each probe. Each binary image was then images and the corresponding sheets were independently
decomposed into two sheets using VCS. The sheets were used in the matching process (i.e., sheet image 1 of all the
superimposed to get the target image. The reconstructed as private images were matched against each other; sheet
well as the original grayscale retina probes were matched image 2 of all the private images were matched against
against the images in the gallery. These experiments each other; reconstructed images of all the private images
suggest the possibility of decomposing and storing retina were matched against each other).The public datasets used
images. Two shares are generated Share1 and Share2 as in this experiments were datasets A. This experiment
output of visual cryptography algorithm. One share along resulted in three EERs: the first was a result of using the
with username is kept by system and other is given on the reconstructed target images for matching, while the
user ID Card. Table 2 shows the result of using the second and the third EERs were a result of using the first
reconstructed retina as probes; the performance is reported sheet and second sheet, respectively, for matching. The
as a function of the different threshold values used to results in Table 3 confirm the difficulty of exposing the
binarize the original probe images. It is observed that a identity of the secret retina image by using the sheets
threshold of 180 results in an EER of 7.05%. alone.
Table 1. Equal Error Rate (%) at different threshold values Table 2.Equal Error Rate (%) for the experiment shows the individual
Threshold Equal Error Rate shift images to reveal the secret images
120 28.2 ERR (%)
150 14.6 Reconstructed vs Reconstructed 2.5
Sheet 1 vs Sheet 1 38.7
180 7.05
Sheet 2 vs Sheet 2 35.4
For authentication user provides share which is on the ID
card. The share extracted from this card is superimposed
with corresponding share that is stored in the database,
generates the retina template image as shown in figure 5.
From this retina template image feature template is
80 IJCSNS International Journal of Computer Science and Network Security, VOL.16 No.9, September 2016

5. Conclusion [12] Monwar, Md Maruf, Marina Gavrilova, and Yingxu Wang.

"A novel fuzzy multimodal information fusion technology
Various approaches adopted by researchers to secure the for human biometric traits identification." In Cognitive
Informatics & Cognitive Computing (ICCI* CC), 2011 10th
raw biometric data and template in database are discussed
IEEE International Conference on, pp. 112-119. IEEE, 2011.
here. This work introduces the possibility of using visual [13] Barkhoda, Wafa, Fardin Akhlaqian, Mehran Deljavan Amiri,
cryptography scheme to provide the privacy to biometric and Mohammad Sadeq Nouroozzadeh. "Retina
data. A method is proposed to store retina template identification based on the pattern of blood vessels using
securely in the database using visual cryptography. In fuzzy logic." EURASIP Journal on Advances in Signal
addition the contribution is here is also providing the Processing 2011, no. 1 (2011): 1-8.
privacy to the retina images private image can only be [14] Blanton, Marina, and Paolo Gasti. "Secure and efficient
created only when both the sheets are present. protocols for iris and fingerprint identification."
Experimental results indicate that by applying visual In Computer Security–ESORICS 2011, pp. 190-209.
Springer Berlin Heidelberg, 2011.
cryptography techniques on retina template for more
[15] Qamber, Sana, Zahar Waheed, and M. Usman Akram.
security, matching performance of iris recognition is "Personal identification system based on vascular pattern of
unaffected with extra layer of authentication. human retina." In Biomedical Engineering Conference
(CIBEC), 2012 Cairo International, pp. 64-67. IEEE, 2012.
REFERENCES
[1] Ross, Arun, and Asem Othman. "Visual cryptography for
biometric privacy."IEEE transactions on information
forensics and security 6, no. 1 (2011): 70-81.
[2] Rajanwar, Shubhangi, Shirish Kumbar, and Akshay Jadhav.
"Visual Cryptography for Biometric Privacy."
[3] Labati, Ruggero Donida, Vincenzo Piuri, and Fabio Scotti.
"Biometric privacy protection: Guidelines and
technologies." In E-Business and Telecommunications, pp.
3-19. Springer Berlin Heidelberg, 2012.
[4] Rathgeb, Christian, and Andreas Uhl. "A survey on
biometric cryptosystems and cancelable
biometrics." EURASIP Journal on Information
Security 2011, no. 1 (2011): 1-25.
[5] Simoens, Koen, Julien Bringer, Hervé Chabanne, and
Stefaan Seys. "A framework for analyzing template security
and privacy in biometric authentication
systems." Information Forensics and Security, IEEE
Transactions on 7, no. 2 (2012): 833-841.
[6] Lai, Lifeng, Siu-Wai Ho, and H. Vincent Poor. "Privacy–
Security Trade-Offs in Biometric Security Systems—Part I:
Single Use Case." Information Forensics and Security,
IEEE Transactions on 6, no. 1 (2011): 122-139.
[7] Hao, Hao, Dinesh Kant Kumar, Behzad Aliahmad, M. Z.
Che Azemin, and R. Kawasaki. "Using color histogram as
the trait of retina biometric." In Biosignals and Biorobotics
Conference (BRC), 2013 ISSNIP, pp. 1-4. IEEE, 2013.
[8] Lajevardi, Seyed Mehdi, Arathi Arakala, Stephen A. Davis,
and Kathy J. Horadam. "Retina verification system based on
biometric graph matching."Image Processing, IEEE
Transactions on 22, no. 9 (2013): 3625-3635.
[9] Hämmerle-Uhl, Jutta, Karl Raab, and Andreas Uhl. "Attack
against robust watermarking-based multimodal biometric
recognition systems." In Biometrics and ID Management,
pp. 25-36. Springer Berlin Heidelberg, 2011.
[10] Bringer, Julien, Hervé Chabanne, and Bruno Kindarji.
"Identification with encrypted biometric data." Security and
Communication Networks 4, no. 5 (2011): 548-562.
[11] Arakala, Arathi, Stephen A. Davis, and Kathy J. Horadam.
"Retina features based on vessel graph substructures."
In Biometrics (IJCB), 2011 International Joint Conference
on, pp. 1-6. IEEE, 2011.