Multi-biometric liveness detection: A new perspective

Abstract

The problem of securing valuable data stored in databases has been of great concern to organizations and individuals alike. The more worrisome is the increasing complexity of fraud perpetration by cyber criminals which demands that a more secure method be deployed. Basic Multi-biometric Authentication System was thought to have sealed the vulnerabilities and escape route from cyber criminals, but emerging attack patterns have proved us wrong. In spite of their benefits, multi-biometric systems also have peculiar challenges especially circumvention of security strategy, that is, how susceptible the system or the presented biometric modality is to spoof attacks and identity fraud. Liveness detection has been applied as an anti-spoofing mechanism to checkmate circumvention, however its application approach has thrown up more vulnerabilities. In this paper, we introduce our work and adopt the Structured Systems Analysis and Design Methodology (SSADM) to assist us understand the weaknesses and propose a solution which integrates liveness detection to halt spoofing of legitimate subjects, and propose a different approach for performing liveness detection in multi-biometric systems that significantly minimizes the probability of circumvention and strengthens the overall security strategy of the authentication process. The expected output of the research is a prototype software for multi-modal biometrics that detects, in a randomized sequence, the absence of liveness and blocks access to critical infrastructure by fraudsters.

Full text

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 26
Multi-biometric Liveness Detection – A New Perspective
Kenneth Okereafor
1
, Oliver Osuagwu
2
, Clement Onime
3
3
1
Nigerian National Health Insurance Scheme (NHIS) , Abuja.
nitelken@yahoo.com
2
Department of Computer Science, Imo State University, Owerri profoliverosuagwu@gmail.com
3
Department of ICT, Abdul Salem International Centre for Theoretical Physics, Trieste Italy onime@ictp.it
Abstract
The problem of securing valuable data stored in databases has been of great concern to organizations and
individuals alike. The more worrisome is the increasing complexity of fraud perpetration by cyber criminals which
demands that a more secure method be deployed. Basic Multi-biometric Authentication System was thought to have
sealed the vulnerabilities and escape route from cyber criminals, but emerging attack patterns have proved us
wrong. In spite of their benefits, multi-biometric systems also have peculiar challenges especially circumvention of
security strategy, that is, how susceptible the system or the presented biometric modality is to spoof attacks and
identity fraud. Liveness detection has been applied as an anti-spoofing mechanism to checkmate circumvention,
however its application approach has thrown up more vulnerabilities. In this paper, we introduce our work and
adopt the Structured Systems Analysis and Design Methodology (SSADM) to assist us understand the weaknesses
and propose a solution which integrates liveness detection to halt spoofing of legitimate subjects, and propose a
different approach for performing liveness detection in multi-biometric systems that significantly minimizes the
probability of circumvention and strengthens the overall security strategy of the authentication process. The
expected output of the research is a prototype software for multi-modal biometrics that detects, in a randomized
sequence, the absence of liveness and blocks access to critical infrastructure by fraudsters.
Keywords: Authentication, biometrics, liveness detection, spoofing, trait.
___________________________________________________________________________________
1.0 Introduction
The growing sophistication of cyber-attacks by cyber
criminals is a global threat that requires a re-definition
and strengthening of the biometric authentication
process in seeking to advance the proper and beneficial
use of biometrics [1]. We are motivated by the idea
that the proper application of appropriate technology
can curtail the rising spate of cyber criminalities
around the globe, specifically by refining the existing
biometric liveness detection process into a more secure
anti-spoofing mechanism. The goal of this research is
to design and develop a software prototype for
enhanced liveness detection, capable of performing
multiple instances of different trait verifications using
alternating traits and modalities from the same person
for each successive instance. In this work, we adopt
multi-mode biometrics using finger, face and voice
modalities.
Human traits that are suitable for biometric purposes in
line with the generic qualities specified by [2] and [3]
are first captured by a sensor to generate an image
which later gets processed through feature extraction
into a template. Biometric templates exist in the form
of electronic data that can be manipulated in similar
ways as any other form of digital data element. Once
the templates are captured into the appropriate
database (DB) or biometric repository, they become
useful for pattern recognition in either the
identification or the verification (authentication) mode.
Given the criticality of biometric templates for
authenticat
ion, it becomes necessary to deploy
adequate all-
round protective mechanisms and systems
to secure them in storage, in process and in transit.
Although the security of some
of the deployed
protective systems is questionable when utilized alone,
integration
with other technologies such as Identity
Based Encryption (IBE), Public Key Infrastructure
(PKI) or digital signatures results in cryptographically
secure applications of biometrics [4]
, which gives a
reasonable guarantee of an encrypted biometric
authentication.
2.0 Securing Biometrics With Cryptography
The concept of encrypted biometrics evolved in the quest
to mitigate the effects of compromised biometric
template. For a system that uses biometric templates for
identification and authentication,
there is the issue of what
to do when a template has been compromised [5]
. For a
mere password or token-
based system, the solution is
straightforward; the user performs a password reset or
gets a new physical token. However, in a case of

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 27
biometric template
compromise, user cannot renew his
biometrics such as grow a new finger or swap to a spare
eyeball. It appears the solution to a compromised
biometric template lies in the application of the revocable
features of biometric templates.
Revocable templates are biometric templates that have
been enhanced through several different cryptographic
methods to allow for the revocation and reissuance of the
existing biometric token without modifying the underlying
biometric [5]. Revocable biometric templates are also
called cancellable biometrics. They are the resulting code
generated when biometric data has been converted into
random strings suitable to apply cryptographic techniques
for security. The extraction is usually done by fuzzy
extractors [6] or secure sketches. Figure 1 below depicts a
typical sequence that generates revocable biometric
templates
.
Figure 1: Sequence for the revocable biometric templates process. Adapted from [5]
Two methods commonly used to create cancellable
biometric
templates are salting and one-way
transformations
[5]. Whereas salting inserts a known set
of fake data into predetermined locations of the template
to disguise it and allow compromised biometric template
to be recovered,
one-way transformations distort the
biometric template in a revocable but irreversible manner
thereby increasing privacy and accuracy. Since it becomes
impracticable to reveal information from the cancellable
biometrics template, the one
-way transformations used to
create them is also know
n as non-invertible transforms [7].
Cancellable biometric templates are essential for biometric
authentication systems (BAS), especially for those
operated under unattended and/or over networked
environments.
3.0
The Liveness Detection (LD) Landscape
Despite the superiority of Biometric Authentication
Systems (BAS) over
passwords and PINs that can be
forgotten or physical tokens that can be damaged,
misplaced or stolen, they are
still not foolproof. Spoofing
(or copy attack) is a fatal threat for BAS
[8], and occurs
when an impostor attempts to mimic the traits
corresponding to legitimately enrolled subjects
[9]. The
ability to detect spoof attempts is a measure of the
performance and security of BAS.
Liveness Detection (LD) is the process of verifying
that the biometric modality presented or rendered before a
biometric verification system for the purpose of capturing
the biometric trait is real and not fake [10]; and that such a
presenter is medically alive [11], and physically present at
the moment of such capture [12]. LD reads claimant’s
physiological signs of life [13]. Biometric circumvention
describes to what extent a biometric system can be fooled
using fraudulent methods [14], and how susceptible the
modality is to spoof attacks [3] and identity fraud [2].
The goal of any anti-spoofing approach is to
strengthen the security of biometric authentication, and at
a basic level, LD is an anti-spoofing mechanism that
attempts to answer questions concerning the originality of
the trait presented before the BAS scanner. In our analysis
though, we identified that the factors that influence the use
and effectiveness of any liveness detection techniques
include (i) ease of trait acquisition, (ii) nature of trait in
view, (iii) tolerable level of intrusiveness, and (iv) duration
of processing. The overall LD goals are better security and
sustenance of a reasonable balance between False Accept
Rate (FAR) and False Reject Rate (FRR) incidents. A well-
applied LD technique should guarantee a FAR low enough
to ward off the possibility of incorrectly authenticating
impostors, and a marginal FRR low enough not to reject
legitimate users.
3.1 Aspects of LD in Focus
Effectively, LD denotes the methods capable of
discriminating real human traits (live or non-live) from

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 28
synthetic counterfeits made by silicon [15], gelatin [16] or
play-doh [17], with the help of appropriate spoof mitigation
algorithms [18]. Checking for signs of vitality involves the
search for, and measurement of, certain intrinsic properties
[11] (such as thermal, optical, mechanical and electrical
quantities), involuntary properties [8] (such as blood flow,
oxygen saturation and pulse rate), and response to external
stimuli (such as eye blinking). These elements must be tested
for verification. In this section, we critically analyse some
measurable quantities (metric) required for running liveness
detection checks on selected traits.
Fingerprint LD
The use of fingerprint recognition for access
control and other uses is becoming increasingly
common due to its security and ease of use [5]. Despite
its broad application,
the existing fingerprint
recognition systems can be easily deceived, for
example, by presenting a well-
duplicated synthetic
finger [19]. A vital question on fingerprint LD is “
how
do we verify that the fingerprint image presented
before a thumb scanner or fingerprint reader is not
an artificial finger or a fake dummy finger fabricated
out of gelatine [16], play-doh [17], silicon [15]
or any
other spoofing tactics [18]
; or molds made out of
latent fingerprints stealthily picked from or left by
legitimate users, or from the dismembered thumb [20]
of the real enrolee?”
In attempting to answer this question,
fingerprint LD tests check for signs of vitality using a
n
analysis of measurement of some or all quantities as
shown in Table 1
Table 1: Quantities evaluated in a fingerprint LD test
SN
Quantity Description
1 Warmth Test for the presence of normal
warmness within acceptable
temperature range for a living human
body.
2 Pulse Test for the presence of pulse on the
finger as evidence of the presence of a
natural heartbeat.
3 Density Test for the pressure tolerance,
elasticity and texture upon contact
with the finger.
4 Haemoglobin
Test for the presence of blood flow.
5 Oxymetry Test for the appropriate saturation of
oxygen in the blood inside the finger.
6 Blood
pressure [15]
Test for the presence of the force
exerted by the heart’s action of
pumping and circulating blood, in
relation to the diameter and elasticity
of the arterial walls within normal
blood pressure range for each given
gender.
6 Spectroscopy
Test for the relative absorption or
reflection or radiation (eg Infra Red
light) on the submitted finger.
7 Perspiration
Test for the presence of secreted sweat
from pores only found in real live
human finger traits.

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 29
Facial print LD
Primarily facial recognition measures the overall
facial structure including distances between eyes,
nose, mouth, and jaw edges [21]. Generally
speaking, there are three ways (also called
replay-attacks [22]) to spoof facial recognition
[23] as follows: (i) photograph of a valid user,
(ii) video of a valid user, and (iii) 3D model of a
valid user. After acquiring the facial image [24],
face recognition processing [23] in BAS involves
four steps:
• Step 1: The face image is enhanced and
segmented.
• Step 2: The face boundary and facial
features are extracted.
• Step 3: The extracted features are matched
against features in the DB.
• Step 4: The classification or recognition of the
user is achieved.
While all four steps are implemented differently
by different vendors [25], a significant question in
facialprint LD is “can we determine, with some degree
of certainty, that the facial image presented before a
biometric facial camera is not a portrait picture of a
legitimate user merely presented as a static paper
photograph, or disguised in a facial mould or a mask; or
a mere screen/video display of the valid user’s picture?”
Table 2 highlights typical quantities measured in a
facialprint LD
Table 2: Quantities evaluated in a Facialprint LD test
SN
Quantity
Description
1 Nodal
geometry
Test for the conformity of the geometry of nodal
points on the face including nose, cheek, jaw,
eye, socket, forehead, etc.
2 Facial
expressio
n
Test for conformity of trait to involuntary
actions and response to stimuli such as smile,
frown, wink, etc.
3 Mouth
moveme
nt
Test for the presence of the natural pattern of
human mouth movement during speech.
4 Eye
blinking
Test for the presence of a sequence that
indicates the pattern of human eye action.
5 Facial
thermogr
am
Test for the presence of radiation only emitted
by a living human face.
Voiceprint LD
The voice recognition system uses the unique
characteristics of the human voice including
measurement of audible frequency, tone, pitch, etc
to distinguish the subject and used for confirmation
of liveness in authentication. Detecting elements of
liveness in the human voice asks the relevant
question: “How can we confirm that the voice
image presented before a voice recognition system
is not a playback of a pre-recorded audio clip, or a
synthesized voice clip of the legitimate user; or
from a physically-present impostor who is
anonymously mimicking the voice of an authentic
user?”
Vein pattern LD
LD in a vein pattern modality essentially checks for
palm vein matching quantities whose measurement
connote the presence of life in the subject including blood
flow, contour synthesis, geometry of fingers, oxymetry,
spectroscopy, pulse rate, blood pressure, etc. A pertinent
question regarding vein pattern LD is “to what extent can
we verify that the hand modality presented before a
vascular pattern reader or hand geometry scanner is from
a valid user and also a living hand naturally attached to a
living human body and not a standalone dismembered part
or from a cadaver?”

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 30
Eye LD for iris and retina patterns
The focus of eye biometrics is basically to
identify vitality signs that show proof of the presence
of a live human eye whose iris and retina show
measurements indicating liveness. The vital question
is ”how do we verify that the eye image presented
before a retina scanner or iris sensor is not faked
with a mimicking contact lens or other eye image
enhancing agents?”
Measurable quantities for detection of real living
iris or retina include a combination of
physiological characteristics and involuntary
actions such hippus movements, eye blinking,
coloration, blood flow, temperature checks, etc.
Keystroke pattern LD
Keystroke liveness check tends to ask the
question “How can we truly confirm that the
keystroke patterns presented before typing
sequence sensor are generated from a real
physical keypad and are coming from the typing
action of a real physical human being and not
from a pattern captured by a key logger attack
tool or simulated by other keystroke pattern
generators?”
3.2 One-Time Password (Otp) Security
Imperatives
OTP was introduced to provide a pseudo
password in form of a one-
off access code to deal with
one of the
major weaknesses of traditional password,
reusability.
An OTP is a password code used to perform
a timed single instance authentication without possibility
of reuse in future transactions. OTP is mostly used by
online payment systems to provide a one-off
password
code which is sent to the user’s email address or phone
number and must be u
sed within a specified limited
timeframe beyond which the OTP expires.
An expired
OTP becomes unusable and a new code must be
generated and used to complete the transactio
n. Apart
from its short-
lived lifespan, a significant security
benefit of an OTP lies in the added association with the
user’s personal telephone number and/or email address.
Figure 2 below shows a sample of an OTP sent as a
Short Message Service (SMS) saf
e token message to a
user’s phone number for use in the authorization of an
online payment transaction
(a) (b)
Figure 2: One Time Password implementations showing (a) 6-
digit token sent via sms, and (b) online payment
authorization portal [26]
where the sent token is entered as secure approval code to complete a pending
transaction.
For example the ComBiom ® Safe ID USB
stick [27] offers a multi-functional token with integrated
biometric authentication that enables physical access
control and logical access control in one token. Figure 3
below shows several hardware tokens that randomly
generate fixed length, short-
lived, unique codes for
access control and authorization of online payment
transactions
.
(a) (b) (c)
Figure 3: Hardware tokens used to generate random 6-digit OTP security codes for (a) access control [27], and
(b), (c) online payment authorization.

West African Journal of Industrial & Academic Research Vol.16 No.1 December 2016 31
To secure the biometric authentication process in a
multi-factor environment, we take advantage of the
widespread use and reliability of OTPs to introduce an
added element of further strength in corroborating
identity and forestalling circumvention of the
authentication process.
2.3 How Significant Is Multi-Biometric (Mb)
Fusion?
No single biometric method to date can guarantee a
100% authentication accuracy and usage by itself. Multi-
biometrics evolved in response to the need to build more
security into BAS. The combination of multiple
biometric sources, modes and more formidable methods
of authentication is referred to as multi-biometric fusion,
and such a system that operates through any of such
combination is often called a multi-biometric system
[10]. MB is the concurrent application of more than one
biometric source, method or other determining factors as
a distinguishing element of authentication.
The uniqueness of the multi-biometric concept lies
in its emphasis on multiple application of variables,
methods or factors as simplified in Table
3.
Table 3: Description of the multi-biometric fusion concept
Multi-
Biometri
c Fusion
Techniq
ue
Description Of
Technique
Example
Multi-
sample
Multiple presentation of a
sample in varying
fashions.
4R Fingers + 4L Fingers + 2
thumbs (4-4-2)
Multi-
mode
OR
Multi-
identifier
Multiple presentation of a
sample from multiple
sources.
Thumb + Face + Voice + …
System
1
+ System
2
+ System
3
+
…
Multi-
system
Multiple application of
different biometric
hardware from different
OEMs assuming vendor
interoperability is
guaranteed.
Example, using the Lumidigm ®
Mercury M301 fingerprint reader
together with the Verifi ® P5100
thumb scanner [28].
PCA
a1
+ ICA
a2
+ LDA
a3
+ … Multi-
algorith
m
Application of multiple
matching algorithms to a
single trait in sequence.
Using different processing
and feature extraction
methods on the same
biometric data.
Example, [29] discusses a face
recognition system that combines
three different global feature
extraction schemes (Principal
Component Analysis (PCA),
Independent Component Analysis
(ICA) and Linear Discriminant
Analysis (LDA).
(Face)
s1
+ (Face)
s2
+ … Multi-
sensor
Processing of similar
samples with multiple
sensors. Multi-sensor
systems employ multiple
sensors to capture a single
biometric trait [9] or
modality of an individual
[10].
Example, a face recognition
system may deploy a 2D camera
to acquire the face image, and an
infrared sensor in conjunction
with a visible-light sensor to
acquire the subsurface
information of a person’s face.
Li + Ri + … Multi-
instance
OR
Multi-
unit
Application of repeated
instances and iterations of
sources. Here the same
modality or trait is
recorded in terms of
multiple instances or parts.
Example, left iris followed by the
right iris of an individual.
Hybrid
model
Concurrent utilization of multiple
fusion techniques.
A mix of many techniques and sources in
one.

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 32
4.0 Identi
fied Problems With Current LD Approach
In general, LD is an embedded function of the biometric
scanner and different manufacturers implement it in
different ways, generally proprietary to each vendor [25],
but the problem lies in the way the liveness detection
check is currently run in multi-biometric systems by many
vendors as a single instance process.
(1) Deficient technique: As far as we know from
available literature, there appears not to be much research
into a single biometric system that performs multiple
simultaneous instances of liveness checks on the same
person using different traits at each instance prior to
authentication. To the best of our knowledge, no such
system has been proposed either. Most biometric
authentication systems are either limited in the number of
instances checked for liveness or are completely unimodal
in nature.
(2) More Vulnerable: The gap introduced by the
deficiency of multiple simultaneous instances of liveness
checks using multiple traits from the same subject has
serious security implications. The risk is that after a smart
attacker has performed reconnaissance, he can launch a
spoof attack targeting only a single liveness detection
technique on a single trait, concentrating all efforts at
achieving this by taking advantage of the system not
having a way of associating each single liveness check of a
person’s trait to another liveness check on a different trait
of the same person for consistency. This security glitch is
too grievous to be ignored by the global Cybersecurity
community.
(3) Intrusiveness: Operationally, the average
biometric user becomes uncomfortable if the trait
acquisition method tends to be too invasive, restrictive,
demanding or time-consuming; for example a theoretical
multi-identifier liveness detection process could require a
user to recite a pre-written text (test for voice liveness),
while holding a pulse meter (test for vein liveness), and
staring at an iris scanner (test for iris liveness) either
simultaneously or in sequence. In the circumstance, and
even where no physical contacts are made with sensors,
many users still develop a natural apathy against the entire
biometric liveness detection process describing it as
grossly intrusive.
(4) Limited Systems Design: A good number of
existing unimodal biometric systems do not have a built-in
liveness detection module and most uninformed users are
equally unaware of the implications of this limitations.
Economic factors top the list of reasons for the acquisition
of low grade systems that are deficient in the liveness
detection component. On the part of the Original
Equipment Manufacturers (OEM)s and vendors,
inadequate Research and
Development (R&D) is a major
factor militating against the design and development of
quality biometric systems with embedded liveness
detection component.
4.1 Proposed Mitigation Approaches
The way and manner, hence the approach, in which LD is
applied i
n a biometric authentication system is significant
to determining the level of security expected and achieved.
Using the Structured Systems Analysis and Design
Methodology, we have thoroughly reviewed existing
liveness detection techniques focusing on thei
r
performance, user acceptance, intrusiveness and security
effectiveness against spoof attacks. We also examined the
comments of various classes of biometric system end-
users and their expectations from future developments.
Based on our analysis of the cur
rent liveness detection
landscape, and having identified its inherent technical and
operational weaknesses, we have developed a new model
of trait vitality checks that is capable of enhancing the
effective security of the biometric authentication strategy
while remaining non-intrusive and user-friendly.
We
present an introductory part of our iterative (recursive)
trait liveness verification model as a series of three
approaches, namely:
• Combination approach
Apply one LD method on a particular biometric tra
it
followed by another dissimilar LD method on a different
trait, from the same enrolee. The rationale of our approach
is based on the fact that physically uncorrelated modalities
or traits (E.g. retina and fingerprint) usually yield stronger
security and
improved performance than correlated
modalities or traits (E.g. lip movement and voice) [9].
First assumption (consistency)
In our model, we assume that at any time during the
authentication process,
LD
count
= T
count
where LD
count
= Number of li
veness
detection instances, and
T
count
= Number of traits prompted for.
The rationale is that, applying n
separate LD methods
(supposing an n-factor multi-
biometric authentication) on
n
separate traits but from the same subject, defeats the
attack purpose since an attacker would naturally be

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 33
expected to perform
n separate spoofs, one for each of the
liveness detection techniques applicable to the particular
trait used or prompted for.
Second assumption (paranoia)
We further assume that the a
ttacker has cleverly produced
all possible spoofs applicable to a particular trait in
readiness to any liveness detection check applicable to his
target trait only.
Therefore prompting for a second, a third, and possibly an
th
different instance of liveness detection using a different
trait for each instance makes it more difficult for the
attacker to successfully circumvent all the options.
The near-intractability of (the attacker) having to spoof
each known liveness detection method for each trait used
in the biometric authentication system, up to the count of
liveness detection instances permissible in the system,
decreases the probability of spoofing, discourages the
attacker and greatly improves the overall system security.
Our approach is illustrated with some tables below
.
Table 4: Our LD approach, instance 1 on fingerprint trait
Modality Human
thumb/finger
Trait 1: Fingerprint
Test of warmth (temperature test).
Test of oxygen saturation in blood
(oxymetry test).
LD checks
applied
Test of sweat secretion from pores
(perspiration test)
LD
Ins
tan
ce
1
Probability
score
P1
Table 5: Our LD approach, instance 2 on facialprint trait
Modality Human
face
Trait 2: Facial
print
Test for instantaneous radiation (facial
thermograph).
Test for effect of background illumination.
Test of light absorption (spectroscopy) on
skin.
Test for effect of variable focus.
Test of eye blinking sequence.
LD checks
applied
Test for natural facial expressions (smile,
frown, etc.)
LD
Ins
tan
ce
2
Probability
score
P2

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 34
Table 6: Our LD approach, instance 3 on iris pattern trait
Modality Human
eye
Trait 3: Iris
pattern
Test of pupil pulsation (Hippus test).
Test of infra-red scattering
LD checks
applied
Aqua reflection density test.
LD
Ins
tan
ce
3
Probability
score
P3
Table 7: Our LD approach, instance 4 on voiceprint trait
Modality Human
voice
Trait 3: Voiceprint
Test for frequency within the audible range.
Test for concurrency with lip movement.
LD checks
applied
Other ancillary tests
LD
Ins
tan
ce
4 Probability
score
P4
Probability
Overall probability of liveness is the mean of P
expressed as a percentage.
P
t
= P
1
+ P
2
+ P
3
+ P
4
… + P
n
(1)
∑ f(p)n/ n
(2)
The probability module built into the LD algorithm
computes the mean matching score based on a
predefined rule-set determined partly by the count of
instances and the security criticality required from the
system in its area of application, which is the basis for
the manual calibration of the system. The system
calibration determines its sensitivity in controlling error
rates.
System tolerance
To reduce the probability of high False Accept
Rates, our system is built to tolerate a low score from
not more than one LD instance per subject.
Randomization approach
By randomizing the choice and sequence of the
possible liveness detection instances through
appropriate algorithm, the attacker faces the
unpredictability of guessing which next trait to expect
and this situation further reduces his chances of beating
the False Accept Rate (FAR) – False Reject Rate (FRR)
balance. Randomization is automated as a built-in
programme module into the BAS to increase overall
security.
By prompting the user for a random set of
traits at the point of acquisition [9], our model shows
that the multi-biometric activates a challenge-response
mechanism, ensuring that the system is interacting with
a live user. Furthermore, to maintain the FAR - FRR
balance (and sustain a zero tolerance for type-2 errors),
the sensitivity of the BAS can be tuned to such a less-
sensitive range that False Accept (FA) possibilities are
significantly reduced without considerably impacting on
False Reject (FR).
Innovativeness of the Proposed Approach
A lot of innovations can be built around this concept of
iterative (recursive) Liveness Detection.
Simultaneity
Firstly, the fusion of the combination and
randomization approaches constitutes a unique iteration,
a sequencing we term recursive liveness detection. Our
approach allows for simultaneous liveness checks on
multiple traits, thereby minimizing delays and reducing
possibility of fatigue-induced user apathy.

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 35
Synchronized processing
Secondly, the BAS can run the matching algorithm
in synch with the trait supply thereby minimizing delays
and overbearing processing time.
5.0 Further Research
Untapped areas exist in MB and LD,
including the need to focus research on developing a
comprehensive taxonomy of LD necessary to advance
further knowledge in the field of biometric securit
6.0 Conclusion
Biometric technology and Biometric Authentication
Systems (BAS) have come to stay, at least going by the
rate of advancing research and development including
innovations in LD techniques. Every LD technique
tends to ask “does the biometric sample being captured
represent an actual measurement from an authorized,
live person?” A negative answer connotes
circumvention, and all known biometric modalities and
traits can be circumvented with varying degrees of ease
irrespective of whether physiological or behavioural.
Although each trait possesses measurable characteristics
that can be used to verify liveness and checkmate
spoofing, it is the application of these characteristics
that makes all the difference. The way and manner,
hence the approach, in which the LD technique itself is
applied within the BAS is significant to determining the
level of security expected and achieved.
In this paper our biometric liveness detection
approach which is based on the appropriate combination
of traits from different uncorrelated modalities of the
same person in a recursive manner has been presented.
The outcome of our study will hopefully assist future
development of anti-spoofing countermeasures not only
to detect and prevent but also to mitigate effects of
successful spoof attacks. The expected Liveness
Detection prototype runs on Oracle Relational Database
Management System (RDBMS) as the backend engine,
the Open Database Connector (ODBC) as Application
Programme Interface (API) and in Java as the front
engine development language. The High Level Model
illustrated in Fig 4 consists of a control centre with the
following automation boundaries: housekeeping,
biometric inputs, analytics module, metric computation
module, report \module
.
Module. etc.
Fig 3.4: High Level Model of the Multi-Moda
l Random Trait Biometric Liveness Detection System
References
[1]
S. Marcel and A. Nouak, “European Association for Biometrics (EAB) establishes new pan
European research collaboration,” European Association for Biometrics, Naarden, The
Netherlands, 2015.
[2] A. K. Jain, “Some Challenges In Bometrics: Facial Sketch, Altered Fingerprints & SMT,”
Idiap Speaker series: Swiss Centre for Biomerics Research and Testing, Martigny

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 36
Switzerland., 2013.
[3] “Biometrics Metrics Report v3.0,” US Military Academy, 2012.
[4] N. D. Sarıer, “Biometric Cryptosystems Authentication, Encryption and Signature for
biometric identities,” Istambul, Turkey, 2011.
[5] Dev Technology, “Emerging Biometric Technology: Revocable Biometric Features,” Dev
Technology Group, 7 Nov 2013. [Online]. Available: http://devtechnology.com/emerging-
biometric-technology-revocable-biometric-features/. [Accessed 27 April 2016].
[6] J. Bringer, H. Chabanne and B. Kindarji, “Identification with Encrypted Biometric Data,” in
Communication on Information Systems Security Symposium, International Conference on
Communications (ICC) 2009, Dresden, Germany, 2009.
[7] K.-H. Cheung, A. Kong, J. You and D. Zhang, “An Analysis on Invertibility of Cancelable
Biometrics based on BioHashing,” Biometrics Research Centre, Department of Computing,
The Hong Kong Polytechnic University, Hung Hom,, Kowloon, Hong Kong..
[8] S. A. C. Schuckers, “Spoofing and Anti-Spoofing Measures,” Elsevier, New York, 2002.
[9] A. Ross, “AN INTRODUCTION TO MULTIBIOMETRICS,” in 15th European Signal
Processing Conference (EUSIPCO 2007), Poznan, Poland, 2007.
[10] H. Li, K.-A. Toh and L. Li, Advanced Topics in Biometrics, World Scientific Publishers,
2103.
[11] G. S. Sawhney, Fundamentals of Biomedical Engineering, New Delhi: New Age
International (P) Ltd., Publishers, 2007.
[12] “Biometrics - Presentation Attack Detection - Part 3: Testing, Reporting and Classification
of Attacks.,” ISO/IEC Standard JTC 1/SC 37 30107-3, 2014.
[13] A. K. Jain, A. Ross and S. Pankanti, “Biometrics: A tool for information security,” IEEE
Transactions on Information Forensics and Security, vol. 1, no. 2, pp. 125 - 143, 2006.
[14] “The impact of biometrics: Technologies of Control,” 2015. [Online]. Available:
https://www.le.ac.uk/oerresources/criminology/msc/unit8/page_19.htm. [Accessed April
2016].
[15] K. M. Valsamma, “Aadhaar, Function Creep and The Emerging Symbiotic Relationship
between Society and Technology,” PARIPEX - Indian Journal Of Research (ISSN - 2250-
1991), vol. 3, no. 8, pp. 184 - 185, 2014.
[16] D. Menotti, G. Chiachia and A. Pinto, “Deep Representations for Iris, Face, and Fingerprint
Spoofing Detection,” IEEE Transactions on InformationForensics and Security, vol. 10, no.
4, pp. 864 - 879, 2015.
[17] D. Gragnaniello, G. Poggi, C. Sansone and L. Verdoliva, “An Investigation of Local
Descriptors for Biometric Spoofing Detection,” IEEE Transactions in Information Forensics
and Security, vol. 10, no. 4, pp. 849 - 863, 2015.
[18] “Spoof Mitigation and liveness detection solutions for the biometric authentication
industry,” 2013. [Online]. Available: http://nexidbiometrics.com/technology/spoof-lab/.
[Accessed April 2016].
[19] E. Marasco and C. Sansone, “Combining perspiration- and morphology-based static features
for fingerprint liveness detection,” Journal of Pattern Recognition Letters, vol. 33, no. 9, p.
1148 –1156, 2012.
[20] A. Abhyankar and S. Schuckers, “Integrating a wavelet based perspiration liveness check
with fingerprint recognition,” Elsevier Ltd., New York, 2007.
[21] O. S. Adeoye, “Multi-mode Biometric Solution for Examination Malpractices in Nigerian
Schools,” International Journal of Computer Application (0975 - 8887), vol. 4, no. 7, pp. 20
- 26, 2010.
[22] I. Chingovska, A. Anjos and S. Marcel, “On the Effectiveness of Local Binary Patterns in

West African Journal of Industrial & Academic Resarch Vol.16 No.1 December 2016 37
Face Anti-spoofing,” IEEE In Proceedings, International Conference of the Biometrics
Special Interest Group (BIOSIG) 2012, September 2012.
[23] G. Pan, Z. Wu and L. Sun, “Liveness Detection for Face Recognition,” Department of
Computer Science, Zhejiang University , China, [Online]. Available:
http://cdn.intechopen.com/pdfs-wm/5896.pdf..
[24] B. G. Nalinakshi and S. M. Hatture, “Liveness Detection Technique for Prevention of Spoof
Attack In Face Recognition System,” International Journal of Emerging Technology and
Advanced Engineering, vol. 3, no. 12, 2013.
[25] D. Simon, “On liveness algorithm fusion.,” Secure Planet, Washington DC. , 2015.
[26] “Online payment engine,” Webpay, [Online]. Available:
http://www.interswitchng.com/#WEBPAY. [Accessed 2015].
[27] “ID All-in-One through USB Token,” Biometric Swiss Identity Security Solutions, 2014.
[28] “Fingerprint Scanner Reviews (Top 10 Reviews),” reviews & comparisons, 2016. [Online].
Available: http://www.toptenreviews.com/computers/scanners/best-fingerprint-scanners/.
[Accessed April 2016 2016].
[29] X. Lu, Y. Wang and A. K. Jain, “Combining Classifiers for Face Recognition,” IEEE
International Conference on Multimedia and Expo (ICME), vol. 3, pp. 13 - 16, 2003.