Content uploaded by Bilgehan Arslan
Author content
All content in this area was uploaded by Bilgehan Arslan on Apr 29, 2018
Content may be subject to copyright.
Security Perspective of Biometric Recognition and
Machine Learning Techniques
Bilgehan Arslan, Ezgi Yorulmaz, Burcin Akca, Seref Sagiroglu
Department of Computer Engineering, Gazi University, Ankara, Turkey
Email: bilgehanarslan@gazi.edu.tr, ezgi.yorulmaz.94@gmail.com, burcinakca06@gmail.com, ss@gazi.edu.tr
Abstract—Biometric systems may be used to create a remote
access model on devices, ensure personal data protection, per-
sonalize and facilitate the access security. Biometric systems
are generally used to increase the security level in addition
to the previous authentication methods and they seen as a
good solution. Biometry occupies an important place between
the areas of daily life of the machine learning. In this study;
the techniques, methods, technologies used in biometric systems
are researched, machine learning techniques used biometric
aplications are investigated for the security perspective, the
advantages and disadvantages that these tecniques provide are
given. The studies in the literature between 2010-2016 years,
used algorithms, technologies, metrics, usage areas, the machine
learning techniques used for different biometric systems such
as face, palm prints, iris, voice, fingerprint recognition are
researched and the studies made are evaluated. The level of
security provided by the use of biometric systems by developed
using machine learning and disadvantages that arise in the use
of these systems are stated in detail in the study. Also, impact
on people of biometric methods in terms of ease of use, security
and usages areas are examined.
Index Terms—Biometric, machine learning, recognition, face,
iris, teeth, fingerprint, voice, security.
I. INTRODUCTION
Biometry considering the contemporary application areas;
it is expressed as a discipline that provides convenience and
speediness in all fields from many perspectives. The last point
of today’s technology shows that; so long as the information,
the information accessibility and the importance level of the
obtained information get be more precise, it is expected that
protection mechanisms have as high as reliability. Because
of that, various mechanisms have been used to prevent the
accessibility of information by everyone. However, when the
availability, practicality, validity, accuracy and security level
that provided of these mechanisms are examined, is observed
that the biometric recognition systems will be located on the
top row. In this context, biometric applications have given
an opportunity that provide to the discrimination to between
people from each other and the definition of unique physical
properties that can be changed person and exhibited behaviors.
Biometric recognition systems show variety according to
characteristic features that used in. In the process of bio-
metric recognition; receiving personal characteristic feature,
analyzing, storage of attributes by decomposing and then,
measurement of the similarities of between those and another
sample that received from same person by comparing each
other are constant [1]. Whichever of biometric feature is
used, these stages are implemented in common, however the
methods and techniques that used in these stages may be
different. In this process, machine learning is located between
the most commonly used method [2]. Because as known, in
the machine learning methods an existing data is taught to
developing system and the accuracy of this learning process
is analyzed with different test data. One of subjects that focus
machine learning is the perception of complex patterns on
computers and the providing rational decisions. In this context,
in biometric applications, machine learning techniques are
used extensively.
The main purpose of using biometric recognition methods is
to create a controlled and secure platform. However, the most
important point in the recognition process using the charac-
teristic is storing the data in top level of safety environment.
Therefore, the security problems that could be encountered in
the implementation phase of the use of biometrics, evaluation
of these application in terms of security, ease of use and usage
areas have been emphasized in section II.
In this study; biometric recognition processes that have
used different biometric features, characteristic features that
provide the highest reliability and using commonly, machine
learning techniques that used in the recognition process,
advantages and disadvantages of these techniques according
to each other and success rates that obtained have been
investigated by offering a general perspective on biometric
systems. For studies conducted between 2010-2016 years, a
research was performed on WOS database. The keywords of
"machine learning", "security" and "biometric recognition" are
queried. In biometric applications that obtained from literature
researches, implementation processes of machine learning
techniques commonly used have been analyzed and machine
learning tecniques have been evaluated for security persperc-
tive in section III. The principle of the biometric recognition
process has been analyzed, working mechanisms of machine
learning techniques used in the biometric recognition process
has been described in detail. Finally, the research results are
discussed in Section IV.
II. MAJOR PROBLEMS IN BIOMETRICS FOR SECURITY
PERSPECTIVE
There have been available a few studies interested in bio-
metric recognition and machine learning techniques in terms
of security perspective. Because biometry is the most common
2016 15th IEEE International Conference on Machine Learning and Applications
978-1-5090-6167-9/16 $31.00 © 2016 IEEE
DOI 10.1109/ICMLA.2016.183
492
tecnique used in recognition process. The methodology of our
study was performed using following steps. These are:
•This research was performed on WOS database studies
conducted between 2010–2016 years.
•The keywords of "machine learning", "security" and "bio-
metric recognition" are queried while scanning studies.
•It was seen that; machine learning techniques are used
not only biometric recognition-authentication process, but
also to provide secure platform for biometric applications
and to control vulnerabilities because of user. For this
reason; biometric recognition process has been analyzed
according to biometric application vulnerabilities, indi-
vidual’s ease of use and usage areas, also appropriate
studies were selected for these criteria.
•Common techniques and methods have been identified
and analyzed using this scan results. Observed results
show that; PCA, LDA, KNN, Naive Bayes, SVM, GMM,
ANN, HMM and LBP are used as machine learning
tecniques for security perspective.
•Four main problems have been identified in the literature
research using keyword of "biometric application" and
"security". These are; user’s session information cap-
turing, maintaining characteristic data, institutional or
organizational vulnerabilities while using biometric appli-
cation and usage of biometric data for profit. Determined
under these four purposes; developed studies examined
and used machine learning techniques, success rates,
usage areas and dataset in this studies have been analyzed
for security perspective.
Machine learning techniques are used not only for extraction
characteristics but also different purposes during biometric
recognition process. Examples of these different purposes are
liveness detection of biometric traits, classifying biometric
data sets to analyze and perform the classification process in
order to improve the statistical approach used in the biometric
authentication process and to increase speed and reliability
in developed biometric methods etc. [7],[12-14], [16], [21-
32]. In this process, it is seen that the biggest weaknesses are
the storage of the biometric characteristics, protection char-
acteristics features of biometric traits [3] and problems that
occur during the using of biometric applications. Therefore,
biometric recognition process has been analyzed according to
biometric application vulnerabilities, individual’s ease of use
and usage areas in this chapter.
The protection techniques and methods we have mentioned
so far, enable the reliable biometric system to be implemented
on the platforms. In addition, in spite of every precaution;
there are some factors that may prevent data confidentiality,
integrity and availability. The violation of this policy is a direct
attack on data and individual privacy of the person. What
can be done if biometric data of the person is compromised
varies with the imagination of who captured. In order to
demonstrate the vialotions of individuals, systems, factors,
methods, tecniques and technologies, we can suggest and
summarize some scenarios that may occur during biometric
recognition process. These are:
•Case I: A user’s session information can be captured.
As a result of the login process using biometric feature,
attacker may have the same rights as the user. It is well
known a person’s private information can be violated such
as first name, last name, address, phone, e-mail or pass-
words, in some cases, even credit card information etc.
Individuals might also get involved crimes unintentionaly
with the use of the private data.
•Case II: The private data might be compromised as a
result of any attack. These data might be shared or sold
for profit or other purposes.
•Case III: Biometric feautes might be collected from the
institutional or organizational vulnerabilities. For exam-
ple, National ;ID numbers have been recently shared in
internet belonging to Turkey, USA, Brazil etc.
•Case IV: Biometric features might be collected from third
parties and used for having other biometric features such
as having personality from face features. [4], [5].
•Case V: Using biometric features for password or other
related applications is dangerous. The reason for that is
the limitation of being unique values. For example, once
a biometric feature is captured, it means that a password
of individual is obtained. So it violates confidentiality and
integrity directly.
As can be seen from the scenerious mentioned above, bio-
metric system have some deficiencies and most important parts
that should be protected for providing to data confidentiality,
integrity and availability with the available tecniques and
technologies.
There have been available a few studies interested in
biometric recognition and machine learning techniques in
terms of security perspective. Because biometry is the most
common tecnique used in recognition process. This research
was performed on WOS database studies conducted between
2010-2016 years. The keywords of "PCA, LDA, KNN, Naive
Bayes, SVM, GMM, ANN, HMM, LBP" with "Security" and
"biometric recognition" are queried. Common techniques and
methods have been identified and analyzed using this scan
results.
III. MACHINE LEARNING SOLUTIONS IN BIOMETRICS FOR
SECURITY PERSPECTIVE
It is seen in these cases that; two key elements must
be considered to develop a secure biometric identification
platform. These are:
•The recorded information during the application can be
stored in a secure manner.
•Application security vulnerabilities that may be encoun-
tered in the usage phase should be minimized.
In this section; the studies have been evaluated with these
five mentioned cases according to application, user, usage
areas and machine learning techniques which is used in
order to improve security of developed applications have been
analyzed.
493
TABLE I
COMPARISON OF USED MACHINE LEARNING TECNIQUES FOR SECURITY PERSPECTIVE,PROPERTIES,DATA SETS,SUCCESS CRITERIA FOR BIOMETRIC
RECOGNITION PROCESS
Ref. Features Security Perspective Developed, used or
proposed methods Success Rate Data Set Case
[7] Face / Palm-
print
Propose a novel cancelable biometric tem-
plate generation algorithm using Gaussian
random vectors and one way modulus hash-
ing
Gaussian Random
Vector /PCA /LDA
Avarage face EER for
PCA 0.05% / Avarage face
EER for LDA 0.03 %
/ Avarage palmprint EER
forLDA0.2%/Avarage
palmprint EER for LDA
0.05 %
ORL/ Yale/ Indian Face/
PolyU/ Casia I, II, V
[12] Face
Present a new technique to protect the face
biometric during recognition, using the so
called cancellable biometric
2DPCA
Improved accuracy of up
to 3% from the original
data
ORL I, II, V
[13] Face
Present a novel biometric protection method
to generate secure facial biometric templates
used in statistical-based recognition algo-
rithms
2DPCA
recognition accuracy by
3% and 4.5% over the
original and other trans-
formed data
ORL I, II, V
[14] Fingerprint Propose a novel binary length-fixed feature
generation method of fingerprint
BCH/
Reed–Solomon/
LDPC
4.58% zeroFAR FVC2002 DB2 I, V
[16] Fingerprint
Focuse on a biometric cryptosystem imple-
mentation and evaluation based on a number
of fingerprint texture descriptors
Gabor Filter/ LBP
EER of the fingerprint
texture descriptors for
FingerCode, LBP8,
LBP16, LBP24, LBPu2,
LBPri, and LDP are
10.96%, 22.79%, 19.54%,
24.6%, 22.88%, 29.56%,
and 15.95%, respectively
FVC2000 DB2a I, II, III,
IV, V
[21] Iris Detect printed-iris attacks / Resist attacks
based on high-quality printing
SVM / LBP / Linear
Kernel / Gabor / So-
bel Filter
FGR 2.25 / FFR 0.25 /
HTER 1.25
Miche Database / Mobio-
fake Database I, V
[22] Mouse
Dynamics
Investigate biometric authentication system
under various different analysis techniques
/ Test static versus dynamic trust models
SVM / ANN / Multi
Classifier Fusion
(MCF) / LibSVM
FMR 0.37% / FNMR
1.12%
Their system based on the
data of 28 users focusing
on different mouse events
I
[23] Face
Growing concerns about security and pri-
vacy, the need to reliably estimate the iden-
tity of an individual has spurred active re-
search in biometrics
SVM with AD / EM
algorithm with NB -VidTIMIT dataset /
MBGC dataset I, II
[24] Face / Fin-
gerprint / Iris
Proposed algorithm continuously updates
the selection process using online learning
SVM / Fusion algo-
rithms FAR 0.01% WVU / LEA I, II, III,
V
[25] Face
Propose a computational approach to human
identification based on the integration of
face and body related soft biometric trait
SVM / Gaussian ker-
nel / SUM / Bayesian
/ Fuzzy Logic
identification rate of 88% ORLAT&T/ Yale/ MUCT IV
[26] Iris
Focus on recognition, and leave the detec-
tion and feature extraction problems in the
background
ANN / SVM FRR average value of
19.80% CASIA-Iris V1 database I, II, III,
IV, V
[27] Fingerprint
Inadequate performance of biometric sys-
tems for the demand of robustness and high
accuracy/ Biometric verification systems are
reliable in ideal environments but can be
very sensitive to real environmental condi-
tions
SVM/ RSVM EER 0.13% FVC2006 datasets I, V
[28] Teeth Improve the recognition accuracy and to
reduce computational complexity PCA/ LDA / EHMM FDR and FRR error rate
of 8.85%
Database Consisting Of
Teeth Images I, V
[29]
Handwriting
/ Gender /
Age
Develop a robust prediction of the writer’s
gender, age range and handedness
SVM / GMM /
FUZZY / SFI
Fuzzy 81.77% GMM
69.75% / SVM 100% /
SFI 85.18%
IAM-1 / IAM-2 / KHATT IV
[30] Face / Teeth
/ Voice
Propose an enhanced multimodal personal
authentication system for mobile device se-
curity / Fuse information obtained from
face, teeth and voice modalities to improve
performance
EHMM / 2D-DCT
/ MFCC / GMM /
KNN/LDA
EER for face-teeth 2.75%
/ face-voice 3.31% / teeth-
voice 4.22% / face 5.09%
/ teeth 7.75 % / voice 8.89
%
1000 biometric traits
database collected via a
smart-phone /20 biometric
traits per 50 persons
I, V
[31] Face / Speak
Investigate the application of existing face
and speaker identification techniques to a
person identification task on a handheld
device
ASR / SVM
ERR for face 6.57% /
Speak 1.54 % / fused 0.64
%
Face and voice data from
35 different people, 100
images and 64 speech
samples
I, V
[32] Face / Voice
Combine real-time face and voice verifi-
cation for better security of personal data
stored on, or accessible from, a mobile
platform
AAM / MFCCs /
GMM
EER for speakers 4.09%
Face 17.45% The BANCA database I, V
494
A. Application Perspective
Biometric features contain high level of distinctiveness.
However, a person forgets his/her ID cards, passwords, token,
etc., but does not have to worry about forgetting or losing
biometric traits. Biometric data cannot be changed. This
situation has both advantages and disadvantages. Passwords
can be cracked easily and can be changed, forgotten or lost if
desired. But biometric data cannot be changed and it is fixed. If
the stored biometric templates seize while recognition or data
transmission process, there is no possibility of changing bio-
metric data. Therefore, it is important to protect the biometric
templates. Lack of protection of these templates or potential
misuse of the stolen templates may rise to huge security
weaknesses such as confidentiality, integrity, availability and
especially privacy issues. In order to prevent weaknesses in
the use of biometric systems, researchers have been working
on some protection mechanisms.
Biometric technologies are important platforms for identity
and privacy assurance. In terms of privacy concerns; the
biometric data can be used in many applications due to the
features having privacy issues. The biometric data is not only
used to take over the session of the person but also can
provide access to private information of individuals such as
identification information, bank accounts, health data, etc.
These features might be used in illegal platforms for illegal
activities if they are not properly secured or used.
The biometric recognition process is performed in two
stages: enrollment and authentication [6]. The identity is
used in two steps for different purposes in the recognition
process. In order to create a record of biometric traits, identity
information is used. Also, identity is used for verifying at the
time of a transaction. Any error occurred in this affects directly
the biometric recognition process.
Several approaches developed against these concerns about
stored biometric template are available. In the literature, pro-
tection methods can be classified into two categories: trans-
formation based approaches and biometric cryptosystems [7]–
[20]. In transformation based approaches, biometric features
reform by using a private key and biometric template saves in
the database with the new form. The matching process can be
carried out in the transformed database during verification [7]–
[13]. This method is reliable because original biometric data
is never recorded in the database. In the studies, feature trans-
formation method have been categorized salting (biohashing)
and non-invertible transform (robust hashing) [6].
The other approach named biometric cryptosystem is used
as a key. In this approach, biometric data is encrypted with
the key but the key is to be reshaped with a different helper
data [14]–[20]. Namely, the key is meaningless without helper
data. The helper data does not include information with keys or
biometric template. At the same time, it can be used to recover
the key when the original biometric is presented in [14]–
[20]. In the studies; biometric cryptosystem method has been
categorized key binding (Fuzzy Vault, Fuzzy Commitment)
and key generation (Secure Sketch Fuzzy Extractor) [6].
Some studies have been made on the conditions mentioned
in section 2, and it has been observed that; the most important
factor in the release of these cases is some attacks may
occur in the application development process such as spoofing,
denial of service, false enrollement, eavesdropping etc. and
weaknesses that may occur as a result of this attacks [12-14].
Therefore; present machine learning techniques in biometric
applications, multiple combinations of these techniques and
improved versions of these methods are used in biometric
applications for measuring the level of security. Machine
learning techniques used in storage process of trait data are
summarized in Table 1.
If biometric template storage process is desired to be
analyzed in terms of security perspective [7],[12-14],[16]:
•Transformation based approaches and biometric cryp-
tosystems are used in order to protect data properly.
•In this study, face and fingerprint data protection have
been examined commonly. Because biometric character-
istic of fingerprint and face is used widely more than the
others.
•Different approaches such as cancelable biometric tem-
plate generation, some new tecnique to protect data
during recognition or some encryption techniques are
focused for creating secure biometric recognition envi-
ronment.
•It is observed in the studies that; PCA, LDA and LBP are
used in machine learning techniques for ensure security
platforms widely.
B. Ease of Use and Usage Areas Perspective
The biometric technology which can make analysis of
characteristic features has also disadvantages as well as ad-
vantages. The biggest disadvantage is the necessity of using
extra equipment during use of this system. For example, during
iris pattern identification high-resolution cameras are required
or for fingerprints recognition, external hardware such as
fingerprint readers are used. Nowadays, because of digitalizing
each transaction increasingly, public or private organizations
feel the need to use these systems in daily lives. Biometric
systems are expensive technologies. Because the methods used
in these systems are expensive and there are limitations in
terms of usage areas. In addition, for operation of the system
there are lots of necessary hardware equipments. Therefore,
the cost of available technologies is a major disadvantage for
biometric systems in use.
Except the hardware costs, another problem in the biometric
systems is to be changed of the measure values. So far, when
the code, token, ID cards and etc. applications that is used in
order to increase the security level are stolen or lost conditions
such as, could eliminate the problem is concerned by contact-
ing the replacement or renovation process. However, if the
biometric data content of person’s characteristics feature such
as fingerprint samples, iris pattern map pass into the hands
unwanted, it cannot be expected to change this property from
the person. Because the person cannot change his fingerprints
495
or hand geometry information, this kind of data must be
protected by strict security measures.
Furthermore, characteristic feature used can cause problems
in the recognition process as well. For example, in the recog-
nition process of the features like face, palm prints, voice and
fingerprints, some variables such as age-related changes in the
facial times, accident, injury, illness, aesthetics etc. may be
caused of the wrong recognition of person. On the other hand,
iris recognition systems are the most valid and accuracy proven
system, it might not be preferred because in technologies high
cost.
Machine learning tecniques are used for many different
purposes if they are evaluated in terms of the user and
usage areas. To increase the obtained success rate in available
biometrics applications, more fast and reliable application
process, tolerate weaknesses that may arise from the user etc.
may be an example for using machine learning techniques
in biometric recognition process. Some studies have been
made on the conditions mentioned in section 2, and it has
been observed that; the most important factors for security
platforms in biometric recognition process are to develop a
system that eliminates the weaknesses users and determine
areas of use appropriately. Machine learning techniques used
for this purpose are shown in Table 1.
If usage area of biometric application and user groups are
desired to be analyzed in terms of security perspective [21-32]:
•The most important topics in the developed application
are to develop machine learning methods which is more
capable of precise measurement and achieve greater suc-
cess in the biometric recognition process.
•The purpose of the examined studies is to eliminate the
problems arising from the user and usage area. Because
of that; some hibrid or advanced machine learning tech-
niques which can detect liveness, resist attacks, test the
application under different various attacks or improve ac-
curacy performance are used to create security platforms
and applications.
IV. CONCLUSIONS AND RECOMMENDATIONS
In this study; machine learning tecniques used biometric
authentication and identification process for security perspec-
tive were examined in detail, recent studies based on machine
learning applied to these process was reviewed.
Regarding the results obtained in the reviewed studies [1-
32]:
•To provide a secure platform for biometric applications
process has two main purposes. These are to protect
biometric characteristic features safely and prevent threats
to occur during the use of biometric applications.
•Machine learning tecniques are used in biometric applica-
tion for two different purposes: recognition-authentication
process and provide secure platform for biometric appli-
cations.
•The biometric recognition-authentication process has two
main stages: enrollment and authentication. The most
important point to be considered is stored characteristic
collected in a secure manner during biometric applica-
tion process. Therefore, it is necessary to protect the
properties obtained in two main stages. Because of this
reason, transformation based approaches and biometric
cryptosystems are developed for protection.
•Biometric recognition-authentication process with ma-
chine learning techniques is performed two phases. First;
the biometric traits are determined to examine and ex-
tract the features. After, collected characteristic features
are used for training the system with machine learning
algorithms i.e. SVM, KNN, NB, ANN etc. These stages
are fixed. Only the methods are changing for performing
each step.
•To provide secure platform for biometric applications
is consisted two main phases: prevention of weakness
caused by the use and prevention of weaknesses that may
arise from the usage area. Machine learning techniques
have been summarized in Table 1 in order to provide two
main phases.
•It is seen in the studies reviewed; there are two basic op-
tions used for biometry. These are statistical approaches
and machine learning techniques. Achievement values
using machine learning methods diversify according to
the combination of selected tecniques. Generally it is
seen that using hybrid techniques in biometric application
are more effective in comparison with single technique
usages.
•Considering by the rate of preference, based on finger-
print recognition system has the most widespread usage
areas for biometric methods. There are multiple reasons
for this situation. First, the fingerprint has high discrimi-
native property. Also the process of obtaining fingerprints
is easy compared to iris, retina, gait recognition. As
well as, to determine the viability of some biometric
features can cause damage to the individual such as retina.
Because of these, fingerprint recognition is widely used
biometric method in terms of reliability.
•It is necessary to have the data set in order to test the
biometric applications. While fingerprint has maximum
sample dataset, gait analysis has limited set of data sam-
ples. One reason for this is the success rate that achieved
by using developed fingerprint biometric applications.
Other contributions of this article are [7],[12-14],[16],[21-
32]:
•It is seen that, SVM, ANN, GMM, KNN, NB, HMM,
LDA, LBP, PCA and fuzzy logic techniques commonly
used for biometric application.
•Machine learning techniques are more preferred than sta-
tistical techniques in biometric authentication and identi-
fication.
•Machine learning techniques are used for different pur-
poses such as recognition, validation, determination of
liveliness, accelerate existing applications, test existing
applications, characteristic extraction, database classifica-
tion. These different areas are also shown in Table 1.
496
•Success rate of the biometric studies conducted with
machine learning techniques are in between 80-99%.
•SVM, NB and ANN techniques are the most common
and powerful machine learning algorithms compared with
other techniques for biometric applications.
•GMM is commonly used for speaker recognition system.
Because it is designed for analyzing distribution of con-
tinuous measurements or features in a biometric system.
•Naive Bayes is capable of precision measurements in
the biometric recognition process because of its sensitive
classification ability.
•LDA offers a high success rate possibility when used to
analyze two-dimensional images.
•Better accuracy rates are achieved with more comprehen-
sive datasets.
•Feature selection process is the most difficult and ef-
fective phase for biometric recognition. Also selecting
features using some statistical methods manually affects
the success rate directly.
•Using hybrid techniques in solutions are more effective
in comparison with single technique usages.
The contribution of this paper is to review the most recent
biometric applications, machine learning solutions according
to the literature and provide secure login process. This study
also contributes the authors who want to study biometric au-
thentication process, identification process and to create a safe
platforms for biometry with machine learning by providing
them a comprehensive analysis of used methods. System and
application developers can also benefit from our conclusions
while developing new software. Finally, at least but not last,
end users should be aware of the recent precaution tecniques to
correct storage of biometric characteristic data and the coun-
termeasures against them by utilizing our recommendations.
REFERENCES
[1] Alçın, M. (2009). Görüntü ˙
I¸sleme Esaslı Parmakizi Do˘
grulama, Marmara
Üniversitesi Fen Bilimleri Enstitüsü, Yüksek Lisans Tezi, ˙
Istanbul, 1-70.
[2] Damousis, I. G., Argyropoulos, S. (2012). Four machine learning
algorithms for biometrics fusion: A comparative study. Applied Com-
putational Intelligence and Soft Computing, 2012, 6.
[3] Galbally, J., Fierrez, J., Ortega-Garcia, J. (2007). Vulnerabilities in
biometric systems: Attacks and recent advances in liveness detection.
Database, 1(3), 4.] In this section, the problems encountered during the
implementation of biometric systems and solutions are transferred.
[4] Eboh, D. E. (2013). Fingerprint patterns in relation to gender and blood
group among students of Delta State University, Abraka, Nigeria. Journal
of experimental and clinical Anatomy, 12(2), 82.
[5] Gornale, S. S., Geetha, C. D., Kruthi, R. (2013). Analysis of fingerprint
image for gender classification using spatial and frequency domain anal-
ysis. American International Journal of Research in Science, Technology,
Engineering and Mathematics, 1(1), 46-50.]
[6] Jain, A. K., Nandakumar, K., Nagar, A. (2008). Biometric template
security. EURASIP Journal on Advances in Signal Processing, 2008,
113.
[7] H. Kaur, P. Khanna, “Gaussian Random Projection Based Non-invertible
Cancelable Biometric Templates,”Procedia Computer Science,vol. 54,
pp. 661–670, 2015.
[8] S.A.H. Nair, P. Aruna, “Comparison of DCT, SVD and BFOA based
multimodal biometric watermarking systems,”Alexandria Engineering
Journal, vol. 54, no. 4, 2015.
[9] F.S. Assaad, G. Serpen, “Transformation Based Score Fusion Algorithm
for Multi-Modal Biometric User Authentication through Ensemble Clas-
sification,”Procedia Computer Science, vol. 61, pp. 410–415, 2015.
[10] G. Bhatnagar, Q.M.J. Wu, “A novel chaos-based secure transmission of
biometric data,”Neurocomputing, vol. 147, pp. 444–455, 2015.
[11] N.K. Ratha, S. Chikkerur, J.H. Connell, “Generating cancelable finger-
print templates,”IEEE Transactions on Pattern Analysis and Machine
Intelligence, vol. 29, no. 4, pp. 561–572, 2007.
[12] M.A. Dabbah, W.L. Woo, S.S. Dlay, “Secure authentication for face
recognition,” in IEEE Symposium on Computational Intelligence in
Image and Signal Processing, CIISP 2007. hskip 1em plus 0.5em minus
0.4em2007, pp. 121–126.
[13] M.A. Dabbah, W.L. Woo, S.S. Dlay, “Appearance-Based Biometric
Recognition: Secure Authentication and Cancellability,” in IEEE 15th
International Conference on Digital Signal Processing. hskip 1em plus
0.5em minus 0.4em2007, pp. 479–482.
[14] P. Li, X. Yang, H. Qiao, K. Cao, E. Liu, J. Tian, “An effective
biometric cryptosystem combining fingerprints with error correction
codes, ”Expert Systems with Applications, vol. 39, no. 7, pp. 6562–6574,
2012.
[15] B. Prasanalakshmi, A. Kannammal, B. Gomathi, “Biometric Cryp-
tosystem Involving Two Traits And Palm Vein As Key,” Procedia
Engineering, vol. 30, pp. 303–310, 2012.
[16] Y. Imamverdiyev, A.B.J. Teoh, J. Kim, “Biometric cryptosystem based
on discretized fingerprint texture descriptors,” Expert Systems with
Applications, vol. 40, no. 5, pp. 1888–1901, 2013.
[17] B. Acharya, M.D. Sharma, S. Tiwari, V.K. Minz, “Privacy protection
of biometric traits using modified hill cipher with involutory key and
robust cryptosystem,” Procedia Computer Science, vol. 2, pp. 242–247,
2010.
[18] A. Nagar, K. Nandakumar, A.K. Jaintitle, “A hybrid biometric cryptosys-
tem for securing fingerprint minutiae templates, ” Pattern Recognition
Letters, vol. 31, no. 8, pp. 733–741,2010.
[19] A.M.P. Canuto, F. Pintro, J.C. Xavier-Junior, “Investigating fusion
approaches in multi-biometric cancellable recognition, ” Expert Systems
with Applications, vol. 40, no. 6, pp. 1971–1980, 2013.
[20] C. Rathgeb, C. Busch, “Cancelable multi-biometrics: Mixing iris-codes
based on adaptive bloom filters,” Computers & Security, vol. 42, pp.1–
12, 2014.
[21] Gragnaniello, D., Sansone, C., Verdoliva, L. (2015). Iris liveness detec-
tion for mobile devices based on local descriptors. Pattern Recognition
Letters, 57, 81-87.
[22] Mondal, S., Bours, P. (2015). A computational approach to the continu-
ous authentication biometric system. Information Sciences, 304, 28-53.
[23] Chakraborty, S., Balasubramanian, V., Panchanathan, S. (2013). Gener-
alized batch mode active learning for face-based biometric recognition.
Pattern Recognition, 46(2), 497-508.
[24] Bharadwaj, S., Bhatt, H. S., Singh, R., Vatsa, M., Noore, A. (2015).
QFuse: Online learning framework for adaptive biometric system. Pat-
tern Recognition, 48(11), 3428-3439.
[25] Arigbabu, O. A., Ahmad, S. M. S., Adnan, W. A. W., Yussof, S. (2015).
Integration of multiple soft biometrics for human identification. Pattern
Recognition Letters, 68, 278-287.
[26] De Marsico, M., Petrosino, A., Ricciardi, S. (2016). Iris Recognition
through Machine Learning Techniques: a Survey. Pattern Recognition
Letters.
[27] Nanni, L., Lumini, A., Ferrara, M., Cappelli, R. (2015). Combining
biometric matchers by means of machine learning and statistical ap-
proaches. Neurocomputing, 149, 526-535.
[28] Kim, D. J., Shin, J. H., Hong, K. S. (2010). Teeth recognition based on
multiple attempts in mobile device. Journal of Network and Computer
Applications, 33(3), 283-292.
[29] Bouadjenek, N., Nemmour, H., Chibani, Y. (2015). Robust soft-
biometrics prediction from off-line handwriting analysis. Applied Soft
Computing.
[30] Kim, D. J., Chung, K. W., Hong, K. S. (2010). Person authentication
using face, teeth and voice modalities for mobile device security. IEEE
Transactions on Consumer Electronics, 56(4), 2678-2685.
[31] Hazen, T. J., Weinstein, E., Park, A. (2003, November). Towards robust
person recognition on handheld devices using face and speaker identifi-
cation technologies. In Proceedings of the 5th international conference
on Multimodal interfaces (pp. 289-292). ACM.
[32] Tresadern, P. A., McCool, C., Poh, N., Matejka, P., Hadid, A., Levy,
C., Marcel, S. (2012). Mobile biometrics (mobio): Joint face and voice
verification for a mobile platform. IEEE pervasive computing, 99.
497
