An elective multibiometric authenticationChelyabinsk state university
,
Chelyabinsk
,
Russia
This work aims to develop an elective multibiometric authentication. The novelty of this work is to develop the principles of distinction and multibiometric authentication, because at the moment there is no such development. Depending on various conditions and factors, including the availability of electronic means and convenience, resistance to attacks and exploits, disease or injury of users can be selected on the basis of biometric authentication of any such biometrics as rhythm password, voice, dynamic signatures and graphics password. The results of the software development based on the new approach are showed. The possible attacks on the developed system are analyzed, and the conclusions and recommendations on defenses from these attacks are submitted.
There are the following problems in single–biometric systems [
1
]:
The developers and researchers of biometrics offer software implementation
based on a single–biometric and single–sensor paradigm without additional tools
and modules [
2
]. It creates problems in usage and operation [
3
]. However, current
trends show a desire to take a different approach, creating a multibiometric
authentication for different areas [
4
]. The main advantage of this approach is
that the security of access can be enhanced [
5
].
Multibiometric system is a system using multiple biometric modalities and
sensors, which can be integrated at various levels and can be used in different
fusions [
6
]. Biometric characteristics are processed by different methods or fusions
in multibiometric systems. The decision can be made on a fused decision rule
to increase reliability. In addition other authentication methods can be used, for
example, PIN–code, password, rhythm of password, tokens.
Multibiometric systems are known to be high security, protection against
spoof attacks and reliability [
4
]. This biometric system may use multiple
biometrics, multiple biometric samples, multiple decision rules, several normalization or
some feature extraction techniques by achieve enhancement of reliability.
However security and reliability of proposed multibiometric systems leads to
additional processing requirements, user inconvenience and privacy issues. Therefore
the development of multibiometric systems is supposed to find a reasonable
compromise between reliability, security, computational costs and user convenience.
This compromise should be found with some automatic or semiautomatic
methods, and this decision should be limited to the dynamic management of security
and reliability. However works are very little attention paid to theory,
architecture, implementation, evaluation of reliability and performance multibiometric
systems that provide dynamically changing the level of security by selecting
different parameters in multibiometric system. In Section 2, different approaches
to the creation of multibiometric systems are presented.
In this paper, an elective multibiometric authentication will mean
multibiometric system where dynamically varying level of security provided by selecting
its various parameters, including selecting a particular biometric characteristics.
The proposed approach for the elective multibiometric authentication will be
described in detail in Section 3.
For example, the managing permissions in an isolated room without
extraneous can be used for authentication of voice, rhythm of password or graphic
recognition. In another case the authentication can be performed based on the
rhythm of password or signature. It can be selected rhythm of password,
signature or graphic recognition for authentication to implement authentication in
mobile or sensor devices. At the checkpoints it can be used signature
authentication. Today’s very urgent task is developing universal modules that implement
managing permissions based on biometric authentication [
7
].
In addition, the managing permission control system based on biometric
authentication has great practical importance and benefits:
– unique, inalienable and inalienability of biometric characteristics;
– difficulties in carrying out the attack on the selective biometric
characteristics;
– independence from the operating system and encoding;
– selectivity in multibiometric authentication;
– possibility of authentication of person due to illness and disability.
The aim of this project was to develop, research and implementation of an
elective multibiometric authentication.
In Section 4, security of multibiometric authentication are presented as the
most important developing aspects.
An multibiometric authentication
Multibiometrics can be used to resolve various aspects of security management
[
4,5
]. Its main aim is enhance the security of the biometric system.
Below you see different approaches to the creation of multibiometric systems
[
1
]:
– multimodal (more than one biometric trait is used for user identification).
– multialgorithmic (multiple different approaches to feature extraction and
matching algorithms are applied to a single biometric trait).
– multiinstance (multiple instances of a single biometric trait are captured).
– multisensor (information of the same biometric obtained from different
sensors are combined for all).
– multisample (multiple samples of a same biometric trait are used for the
enrollment and recognition).
Multimodal biometric systems can operate in three different modes [
1
]:
– Serial Mode (cascade mode) — each modality is examined before the next
modality is investigated.
– Parallel Mode — sensed/captured data from multiple modalities are used
in concurrent way to perform recognition, then the results are combined to
make final decision.
– Hierarchical Mode — individual classifiers are combined in a hierarchy or
tree structure.
There are the following different levels of fusion in multibiometric system:
decision, score, feature, and sample. Universal system should take into account
all possible approaches to implementation multibiometrics by using fusion [
8
].
There are three strategies for multibiometric fusion [
9
]:
– User-specific normalization for multibiometric fusion. For example,
depending on the quality of input samples, the proposed algorithm intelligently
selects appropriate fusion algorithm for optimal performance [
10
].
– Robustness criterion to rank users according to their performance. It gives
consistently good performance across different databases despite the lack of
training samples. Fisher-ratio, F-ratio, and d-prime reported as examples of
criteria in [
9
].
– Selective fusion strategy. Because not all biometric characteristics need to
be operational for each transaction or the participating biometric systems
can operate independently of each other, we should dynamically select
appropriate fusion algorithm for effective performance.
In the work [
11
] a dynamic score level fusion scheme for a multialgorithmic
recognition by incorporating quality as an input for fusion was investigated.
Smartness has been very tactfully administered to the processing by employing
different efficient algorithms for a given modality. Selection of the recognition
algorithms is rooted on the attributes of the input. If one sensor is not functional,
others contribute to the system making it fault-tolerant. Multiplicity has been
employed to establish a unanimous decision. Information fusion at various levels
has been introduced. Sensor level fusion, local decision level fusion at algorithmic
level and global decision level fusion provide the right inference. A multitude of
decisions are fused locally to decide the weightage for the particular modality.
Algorithms are tagged with weights based on their recognition accuracy. Weights
are assigned to sensors based on their identification accuracy. Adaptability is
incorporated by modifying the weights based on the environmental conditions.
All local decisions are then combined to result in a global decision about the
person.
In the work [
12
] authors propose the design of a sequential fusion technique
that uses the likelihood ratio test-statistic in conjunction with a support vector
machine classifier to account for errors in the former; and the design of a dynamic
selection algorithm that unifies the constituent classifiers and fusion schemes in
order to optimize both verification accuracy and computational cost. Depending
on the quality of the input biometric data, the proposed algorithm dynamically
selects between various classifiers and fusion rules to recognize an individual [
12
].
The resulting algorithms are used to reduce the effect of covariate factors in face
recognition by combining the match scores obtained from two face recognition
algorithms.
In the paper [
13
] presents techniques for performing multibiometric fusion at
the rank level. The proposed methods are suggested to enhance the performance
of rank-level fusion schemes in the presence of weak classifiers or low quality input
images. It’s not required an additional training phase, making them suitable for
a wide variety of databases. Also it should be included performing a comparative
study on the effect of input image quality on score level, rank level and decision
level fusion; using quality factor to select the best probe image for fusion; and
conducting experiments using other databases consisting of different modalities.
Multibiometric systems must be highly flexible to take into account the
different requirements and limitations of users. The system should solve the problem
lack of biometric characteristics, as a result of poor quality or physical problems,
with possibility to use other available biometric characteristics. In addition, it is
important to comply with the requirement necessary security level. It requires
developing a dynamic elective different rules and methods of multibiometric
fusion.
One of the approaches described in article [
14
], which experimented with a
few simple methods of fusion multibiometric.
The authors [
15
] proposed another interesting approach, that includes
conducting continuous authentication. This approach requires a long physical
presence of user and therefore it isn’t suitable for some kinds of applications.
In article [
16
] proposed to use multiple security levels for multibiometric
authentication with three biometric characteristics (face, lip movement, voice).
When the required security level is low, it is sufficient to take a decision on
the basis of two of the three biometric characteristics. On the other side, for
applications with a high security level, this system requires the use of all three
biometric characteristics. However, this system does not provide a way to change
the dynamic security level. Instead, the administrator makes a decision witch
strategies and methods of fusion has to be used.
Interesting architecture for dynamic security management of multibiometrics
has been discussed in [
17
]. This work suggests a scenario of managing permission
in the building with divided into different zones (this can be different floors or
room numbers), and defined access rights for each user. Access solutions in a
particular area may also depend on the solutions already adopted in the other
zones. Furthermore, the amount of biometric characteristics required in each
zone and different elective rules of fusion can be varied.
Another aspect of the development of elective multibiometric system is to
provide the desired performance, as well as the performance of users’ preferences,
constraints, user convenience, and age–related changes [
18
]. Research challenges
of these problems related to the dynamic fusion techniques.
Security level of multibiometric system must also be adjusted depending
on the possible future attacks. This system requires the elective appropriate
methods for the fusion.
In the work [
19
] a new approach for the adaptive combination of multiple
biometrics to dynamically ensure the desired level of security is presented. The
proposed method uses a hybrid particle swarm optimization to achieve adaptive
combination of multiple biometrics from their matching score performance.
Experimental results suggest that the dynamic selection of fusion rules and their
parameters using the proposed method can offer better performance than the
decision level scheme. The work [
19
] is focused to estimate on the performance
improvement. One of the key problems in adaptive multibiometric management
pertains to the selection of biometric modalities.
Therefore this paper is focused to develop algorithms that can adaptively
select best set of biometric modalities from the available set to ensure the desired
level of security.
3
Proposed elective multibiometric authentication
In this paper, as opposed to all previous work it offers a combined approach to
the development of elective multibiometric authentication system that uses all
of the above criteria for selecting the method of fusion multibiometrics.
The criteria for the election of multibiometric authentication is described
as a scheme in Figure 1, which shows the main stages of the elective
multibiometric authentication based on rhythm of password, voice, dynamic signature
and graphical recognition. This approach and design can be generalized to other
biometric characteristics.
The most important building block for this scheme is a unit of
semiautomatic settings, which performs the convert of all settings and parameters set
by the administrator and the user at the stage of training. The parameters and
settings of semi–automatic selection performs order of biometric characteristics,
Keyboard (rhythm
of password)
Microphone
(voice)
set of biometric characteristics, the input devices (sensor), the feature
extraction methods, the matching methods, the method of the score combination and
decision. The choice of semi–automatic selection of the methods is a fusion of
predefined strict rules and criteria.
Here are the basic criteria and rules:
1. Availability of necessary input devices (sensors);
2. The security level (the number of required biometrics);
3. Elective priority of biometrics;
4. The result of previous authentication attempts;
5. Features of area (room, equipment);
6. Features of users and their preferences, age limits;
7. The request time for authentication;
8. The extent of the attacks and spoof attacks on the sensor;
9. The quality of biometric samples.
After setting all parameters, unit semi–automatic settings may select a
desired decision in the block score combination f1(m1; m2; m3); :::; fk(m1; :::; m4)
and a decision threshold, where m1; m2; m3; m4 — result matching each
biometric characteristics individually.
However, this elective multibiometric system doesn’t automatically select the
parameters to guarantee a certain security level; this work is to further research
and development.
In our proposed elective multibiometric authentication there are 4 biometric
characteristics (voice, dynamic signature, rhythm of password, graphical
recognition). All possible subset f1; 2; 3; 4g can be: f g; f1g; f2g; f3g; f4g; f1; 2g; f1; 3g;
f1; 4g; f2; 3g; f2; 4g; f3; 4g; f1; 2; 3g; f1; 2; 4g; f1; 3; 4g; f2; 3; 4g; f1; 2; 3; 4g.
Each of 16 subsets describes one of the choices of biometric characteristics in
elective multibiometric authentication. We describe the algorithm of select
combination of biometric characteristics, depending on the level of security through
the following model.
Let the set used biometric characteristics defined as fp1; p2; p3; p4g, where pi
— the index of using biometric characteristics i. We assume that the criteria
influencing for pi are independent. Then
pi =
k
∏ pij ;
j=1
where pij — assessment of factor of using biometric characteristics i using the
criteria j.
Here are the criteria j for each biometric characteristics i in the proposed
elective multibiometric authentication:
– p1 = f0; 1g — factor of availability of necessary input sensors. p1 = 1 when
input sensor is available, and p1 = 0 when input sensor is not available.
– p2 = [1; 10] — factor of necessary security level. Administrator sets this
factor for each biometric characteristics i. For example, p2 = 10 for implicit
voice authentication, for other biometric characteristic (dynamic signature,
rhythm of password, graphical recognition) p2 = 3; 9; 6 respectively.
– p3 = [1; 10] — factor of using attacks on sensor. Administrator sets this
probability for each biometric characteristics i. For example, p3 = 3 for
voices because of high risk of spoof attacks, for other biometric characteristic
(dynamic signature, rhythm of password, graphical recognition) p3 = 9; 7; 6
respectively.
– p4 = [0; 1] — factor of quality of biometric samples. Depending on the quality
of input samples, the proposed algorithm dynamically selects appropriate
fusion algorithm for optimal performance [
10,11,13
].
– p5 = [3; 10] — factor of result of previous authentication attempts. This
factor dynamically estimate. For example, p5 = d if the last d attempts had
failed to authenticate.
– p6 = [1; 10] — factor of security level of area (room, equipment).
Administrator sets this factor for each biometric characteristics i.
– p7 = [0; 1] — factor of user preferences. Administrator sets this factor for
each user. For example, because of age limits or lack biometric characteristic
then p7 = 0, otherwise p7 = 1.
– p8 = [0; 10] — factor of request time for authentication. For example, p8 = 3
for voices because of long process, for other biometric characteristic
(dynamic signature, rhythm of password, graphical recognition) p8 = 9; 7; 6
respectively.
An algorithm for selecting a subset of elements for elective multibiometric
authentication:
1. Consider fp1; p2; p3; p4g, all values by evaluating all criteria pj .
2. Compare pi with a threshold > 0. If pi < then exclude pi. Administrator
sets this threshold .
3. Once the values have been calculated fp1; p2; p3; p4g, we sort pi desc.
4. Choose the t first pi, which correspond to the high indices selected biometric
characteristic. Administrator sets this parameter t.
In our proposed elective multibiometric authentication, depending on
various conditions and factors, including the availability of electronic means and
convenience, resistance to attacks and exploits, disease or injury of users can
be selected on the basis of biometric authentication of any such biometrics as
rhythm password, voice, dynamic signatures and graphics password.
4
Security of elective multibiometric authentication system
Implementation of biometric systems has problems in the security, so let’s
consider the most important developing aspects of elective multibiometric
authentication [
20,21,22
].
2
4
5
6
Biometric input (sensor) 1Feature extractionMatchingApplication 3Enrolment Database of templates 7
11
8
9
10
The article gives an overview of current attacks and protection measures
[
23,24
]. Here are all the typical attacks, threats related to the elements of the
authentication system (see Fig. 5):
1. An attack on the biometric input (sensor);
2. An attack on the communication channel between sensor and biometric
systems;
3. An attack on the feature extraction;
4. An attack on the communication channel of feature extraction;
5. An attack on the matching;
6. An attack on the score;
7. An attack on the link with the database of templates;
8. An attack on the enrolment;
9. An attack on the channel between the enrolment and database of templates;
10. An attack on the templates;
11. An attack on the application.
All of the above attacks, excepts an attack on the input device, are common
to all biometric systems. Protection against these attacks is to use digital
encoding, timestamp, encrypt the data channel, the special methods to prevent the
introduction of malicious code, antivirus methods and other methods to protect
the information [
25
].
Most interesting attack is an attack on the input sensor in a multibiometric
system, since this attack poses a real threat [
26,27,28,29
]. This attack is directed
to an biometric input (sensor), and occurs when an attacker provides illegitimate
biometric sensor. This attack can be divided into three types:
– forced attack is providing biometric characteristics on illegitimate grounds,
such as the use of violence;
– simulation attack is simulating biometric characteristics by synthesized
biometric data;
– replay attack is replaying previously recorded biometric data.
Many of problems and attacks can be prevented by using digital coding,
timestamp and encrypt data channel. In other words, there are special
cryptographic protocols to help prevent various attacks [
30
].
Also you can use the following methods to prevent attacks:
– Use the methods of detecting the liveness of biometric characteristics;
– Apply the different approaches of organization of database templates and
the structure of template to improve security systems;
– Use multifactor authentication to improve the reliability of biometric
systems;
– Use special methods of reducing biometrics and «encryption personality» to
resolve the problem of confidentiality and protection of biometric data.
The analysis of all threats of the elective multibiometric authentication
system allows making the conclusion: the using of multibiometric and the principle
of selectivity increase the reliability and security, since the attacker must take
into account all the parameters and characteristics of the implementation of
security system and the criteria for selecting all parameters.
Quantitative estimates of reliability and security of multibiometric
authentication doesn’t restrict to test a large base of multibiometric samples, and the
results obtained by generalized theoretical estimates of reliability can be found
in [
5
].
Conclusion
As a result of this work it has be developed the elective multibiometric
authentication system. In this paper, as opposed to all previous work it was offered
a combined approach to the development of elective multibiometric
authentication. In this approach uses different criteria for the choice of semi–automatic
method of fusion and other parameters of multibiometric authentication system.
In addition, an analysis of possible attacks, recommendations of protects and
conclusions were carried out.
However, there are some trends in future developments of the system:
providing greater versatility, using other biometric features, increasing performance
and reliability, implementation of dynamic selection of parameters, in particular,
the method of fusion multibiometric data for a guaranteed level of security.
Acknowledgements. This research was supported by grant RFBR
14-0731049-mol-a.
1.
Gad
,
R.
,
El-Fishawy
,
N.
,
El-Sayed
,
A.
,
Zorkany
, M.:
Multi-Biometric Systems
:
A State of the Art Survey
and Research Directions.
International Journal of Advanced Computer Science and Applications6
(
6
),
128
-
138
(
2015
)
2.
Bolle
,
R. M.
,
Connell
,
J. H.
,
Pankanti
,
S.
,
Ratha
,
N. K.
,
Senior
,
A. W.
: Guide to biometrics. Springer-Verlag, New-York (
2003
)
3.
Dunstone
,
T.
,
Yager
, N.:
Biometric system and data analysis: design, evaluation, and data mining
. Springer, Boston, Ma (
2009
)
4.
Ross
,
A. A.
,
Nandakumar
,
K.
,
JainA. K.
: Handbook of multibiometrics. Springer, New York (
2006
)
5.
Bhanu
,
B.
,
Govindaraju
,
V.
:
Multibiometrics for Human Identification
. Cambridge University Press, Cambridge (
2011
)
6.
Sesin
,
E. M.
,
Belov
,
V. M.:Personal identification system based on integration organization of several biometric characteristics of the person
.
Proceedings of Tomsk State University of Control Systems and Radioelectronics2
(
25
),
2
,
175
-
179
(
2012
)
7.
Ruchay
,
A. N.
:
Development of a universal set of modules for managing permissions windows xp on based biometric authentication
.
Security of information technology 2
,
74
-
78
(
2013
)
8.
Vokhmintsev
,
A.
,
Makovetskii
,
A.
,
Kober
,
V.
,
Sochenkov
,
I.
,
Kuznetsov
, V.:
A fusion algorithm for building three-dimensional maps
.
SPIE Proceedings (9599)
,
1
-
7
(
2015
)
9.
Poh
,
N.
,
Ross
,
A.
,
Lee
,
W.
,
Kittler
,
J.:A user-specific and selective multimodal biometric fusion strategy by ranking subjects
.
Pattern Recognition46
,
3341
--
3357
(
2013
)
10.
Vatsa
,
M.
,
Singh
,
R.
,
Noore
,
A.
:
Context Switching Algorithm for Selective Multibiometric Fusion
.
Pattern Recognition and Machine Intelligence
(
5909
), Springer,
452
-
457
(
2009
)
11.
Fathima
,
A.
,
Vasuhi
,
S.
,
Treesa
,
T.
,
Babu
,
N.T.
,
Vaidehi
,
V.
:
Person Authentication System with Quality Analysis of Multimodal Biometrics
.
WSEAS transactions on information science and applications
12.
Vatsa
,
M.
,
Singh
,
R.
,
Noore
,
A.
,
Ross
,
A.
:
On the Dynamic Selection of Biometric Fusion Algorithms
.
IEEE transactions on information forensics and security 5
(
3
),
470
-
479
(
2010
)
13.
Abaza
,
A.
,
Ross
,
A.
:
Quality Based Rank-Level Fusion in Multibiometric Systems
.
Proc. of 3rd IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS)
(
2009
)
14.
Kittler
,
J.
,
Hatef
,
M.
,
Duin
,
R. P. W.
,
Matas
, J.:
On combining classifiers
.
IEEE Trans. Patt. Anal. Machine Intell
.
20
,
226
-
239
(
1998
)
15.
Sim
,
T.
,
Zhang
,
S.
,
Janakiraman
,
R.
,
Kumar
,
S.
:
Continuous verification using multimodal biometrics
.
IEEE Trans. Patt. Anal. Machine Intell
.
29
(
4
),
687
-
700
(
2007
)
16.
Frischholz
,
R. W.
,
Deickmann
, U.:
BioID: a multimodal biometric identification system
.
IEEE Comput
.
33
(
2
) (
2000
)
17.
Bradlow
,
E. T.
,
Everson
,
P. J.:Bayesian inference for the beta-binomial distribution via polynomial expansions
.
J. Comput. Graphical Statistics11
(
1
),
200
-
207
(
2002
)
18.
Poh
,
N.
,
Wong
,
R.
,
Kittler
,
J.
,
Roli
,
F.
:
Challenges and research directions for adaptive biometric recognition systems
.
Proc. ICB
, Alghero,
Italy.
(
2009
)
19.
Kumar
,
A.
,
Kanhangad
,
V.Zhang
, D.:
A new framework for adaptive multimodal biometrics management
.
IEEE Transactions on Information Forensics and Security. (5)
,
92
-
102
(
2010
)
20.
Ruchay
,
A. N.
:
Prototype centralized system of election multifactor biometric authentication
.
Security of information technology 1
,
118
-
120
(
2013
)
21.
Ruchay
,
A. N.
,
Gorshenin
,
V. V.
,
MatkinI. A.
:
Development of prototype centralized system of elective multifactor biometric authentication
.
Intelligent information processing: theses of the 10th international conference
, Greece,
Crete 2014
.
231
(
2014
)
22.
Wayman
, J.:
Technical testing and evaluation of biometric devices. Biometrics - personal identication in networked society
. Kluwer Academic Publisher.
345
-
368
(
2002
)
23.
Ruchay
,
A. N.:The model of attacks and protection of the speaker recognition biometric system
.
Proceedings of Tomsk State University of Control Systems and Radioelectronics1
(
23
),
96
-
100
(
2011
)
24.
Roberts
,
C.
:
Biometric attack vectors and defences
.
Computers and Security26
(
1
),
14
-
25
(
2007
)
25.
Ruchay
,
A. N.
:
Dependent speaker verification
.
LAP LAMBERT Academic Publishing, Saarbrucken
(
2012
)
26.
Ratha
,
N. K.
,
Connell
,
J. H.
,
Bolle
,
R. M.:Enhancing security and privacy in biometrics-based authentication systems
.
IBM Systems Journal40
(
3
),
614
-
634
(
2001
)
27.
Yanushkevich
,
S.
,
Shmerko
,
V.
,
Stoica
,
A.
,
Popel
,
D.
: Inverse Problems of Biometrics. CRC Press, Taylor and Francis Group (
2005
)
28.
Yanushkevich
,
S.
,
Wang
,
P.
,
Gavrilova
,
M.
,
Srihari
,
S.
:
Image pattern recognition: synthesis and analysis in biometrics
.
Series in Machine Perception and Artifical Intelligence67
(
2007
)
29.
Uludag
,
U.
,
Jain
,
A.
:
Attacks on biometric systems: a case study in fingerprints Proc. SPIE
.
5306
,
622
-
633
(
2004
)
30.
Schneier
,
B.
:
The uses and abuses of biometrics
.
Communications of the ACM42
(
8
) (
1999
)