Protecting Digital Fingerprint in Automated Fingerprint
Identification System using Local Binary Pattern Operator
K. Ait sadi
1
, I. Bouchair
1
, K. Zebbiche
2
and M. Laadjel
2
1
Centre de Développement des Technologies Avancées, Division Architecture des Systèmes,
Cité 20 Août 1956, BP 17, Baba Hassen, 16303, Algier, Algeria
2
Centre de Recherche et Développement de la Gendarmerie Nationale, (CRD-GN), Algier, Algeria
Keywords: Fingerprint, LBP, Arnold Scrambling, Watermarking, AFIS System.
Abstract: The Local binary pattern (LBP) operators, which measure the local contrast within a pixel's neighbourhood,
have been successfully applied to texture analysis, face recognition, and image retrieval. In the paper, we
present a new application of the LBP operators for securing digital fingerprint in AFIS System (Automated
Fingerprint Identification System), while inserting a robust watermark (ID image and Face image) to
increase not only security but also to facilitate the recognition of the person. To improve the security of the
embedding, the watermarks are scrambled using Arnold technique and are then hidden in the fingerprint
image of the corresponding person. Experimental results show that the proposed watermarking method is
robust against commonly-used image processing operations, such as additive noise, luminance change,
contrast enhancement, and JPEG compression while does not change the fingerprint features and maintains
a good visibility of the original fingerprint images.
1 INTRODUCTION
Biometric systems based on fingerprints as AFIS
systems developed by the Federal Bureau of
Investigation (FBI) and other agencies and
researchers aim to identify persons from their
fingerprints previously acquired and stored in the
data base system. However, Due to their popular use
in many applications, these systems are not immune
from errors and attacks that attempt to exploit
vulnerabilities to destabilize their performance.
Uludag, et al. (Uludag and
Jain, 2004)
identify eight
basic sources of attacks that are possible in a generic
biometric system (Figure 1). In the first type of
attack, a false biometric (such as a fake finger) is
presented at the sensor. Resubmission of digitally
stored biometric data constitutes the second type of
attack. In the third type of attack, the feature detector
could be forced to produce feature values chosen by
the attacker, instead of the actual values generated
from the data obtained from the sensor. In the fourth
type of attack, the features extracted using the data
obtained from the sensor are replaced with a
synthetic feature set. In the fifth type of attack, the
matcher component could be attacked to produce
high or low matching scores, regardless of the input
feature set. Attack on the templates stored in
databases is the sixth type of attack. In the seventh
type of attack, the channel between the database and
matcher could be compromised to alter transferred
template information. The final type of attack
includes altering the matching result itself. All of
these attacks have the possibility to decrease the
credibility of a biometric system. Several techniques
based on digital watermarking and data hiding have
been proposed in the literature to enhance biometric
security against the aforementioned attacks.
Figure 1: The different points of attack on the AFIS
system (Uludag and Jain, 2004).
33
Karima A., Imène B., Zebbiche K. and Laadjel M..
Protecting Digital Fingerprint in Automated Fingerprint Identification System using Local Binary Pattern Operator.
DOI: 10.5220/0005055600330039
In Proceedings of the 11th International Conference on Signal Processing and Multimedia Applications (SIGMAP-2014), pages 33-39
ISBN: 978-989-758-046-8
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
Pravin et al. in (Pravin et al., 2011) proposed a
stegano-crysto system for enhancing biometric
feature security with RSA and data hiding. The aim
is to increase security of biometric system and
facilitate identification. They used a biometric code
generated from captured biometric image. The light
of this paper is the use of the RSA algorithm and
data hiding process to protect the biometric
information against attacks. But the authors did not
precise whose is the biometric image used as host
image and those used as watermarks. Also the
authors did not investigate about the robustness of
the data hiding against attacks and distortion
inducted with watermark embedding. Thus, because
sometimes the embedding destroys the biometric
features of biometric image itself. For example,
when the face features are embedded into fingerprint
image, the features of fingerprint may get disturbed
and wrong minutiae points may arise. Another point
important is that the authors did not discuss the
credibility of the verification and recognition
performance under attacks
Jain et al. (Jain and Uludag,
2003)
have hidden
fingerprint image features in face image. Then they
proposed to hide the facial information as watermark
to authenticate the fingerprint image. A bit stream of
eigenface coefficients is embedded into selected
fingerprint image pixels using a randomly generated
secret key. The extraction bits are then employed for
fusion recognition with host fingerprint image.
However, since the Extracted pattern is given for
identification without credibility verification, it only
increases recognition performance under attack free
circumstances thus provide no additional security.
Another important criterion to respect in the data
hiding is the payload which is low in the proposed
algorithm.
Mathivadhani et al. (Mathivadhani and
Meena, 2010)
have presented a comparative study on fingerprint
protection using watermarking techniques.
However, most of these works induced distortion
with watermark embedding sometimes destroys the
biometric features of the fingerprint image itself. For
example, when the face features are embedded into
fingerprint image, the features of fingerprint may get
disturbed and wrong minutiae points may arise.
In addition of these attacks, and by analyzing the
fingerprints based identification system, we found
that generally the enrolled fingerprint images are
stored in a database along with the demographic text
data of the individual and a facial image. The
different data types are usually stored under three
different sub categories in a database. The
collection, storage and analysis of disparate
information introduces problems such as data
mismatches and mishandling, high cost of storage, a
longer time for retrieval, and unauthorized
tampering of the files in the database (Noore et al.,
2007). This leads to the decrease of the biometric
system credibility (Figure 2). Ensuring the security
and the integrity of biometric data has become a
critical issue.
Figure 2: The attack after enrolment in AFIS system.
In order to solve this problem, this paper proposes
an algorithm that combines a digital watermarking
technique and an Arnold scrambling process. The
digital watermarking approach is used to embed the
iconic digital information (watermark) where in our
case; it’s composed of ID image and gray scale face
image into the corresponding enrolled digital
fingerprint of the person. Embedding the facial and
ID data into the individuals fingerprint image
eliminates data mismatch, reduces the high cost of
storage, speeds the retrieval of related data and
detects tampering. It is important to ensure that the
embedded ID and face watermarks do not alter the
functional integrity of the fingerprint and its ability
to detect possible matches.
The unique robustness and security character of
the watermark can ensure the integrity and reliability
of the fingerprint data after information exchange
process. So it can identify the authenticity of the
contents, as well as content protection.
The proposed watermarking method is based on
Shih approach (Shih and al., 2011), where he
proposed a semi fragile spatial watermarking based
on Local Binary Pattern operator (LBP) to embed a
binary watermark. The operator takes a local
neighborhood around each pixel, thresholds the
pixels of the neighborhood at the value of the central
pixel and uses the resulting binary-valued image
patch as a local image descriptor.
In the proposed algorithm, the embedding
process is performed on two levels. In the first level
the binary ID image is inserted while the gray scale
face image is embedded in the second level. To
SIGMAP2014-InternationalConferenceonSignalProcessingandMultimediaApplications
34
improve the security of the embedding; the proposed
technique employs the Arnold scrambling to pre-
process on the watermark images, such that the
watermarked fingerprint image has security under
cryptography sense, for double protection.
2 THE PROPOSED METHOD
2.1 Arnold Scrambling Transform
(AST)
Scrambling transformation as a means of encrypted
technology (yothish et al., 2012) is applied as
preprocess stage of the watermarking, after
scrambling transformation, one meaningful
watermarking will become a meaningless, chaotic
image. If the scrambling algorithm and keys are not
known, the attacker cannot recover it even if he gets
the watermark from the embedded watermarking.
Thus the Scrambling transformation plays a role of
secondary encryption. Additionally, after scrambling
transformation, it will upset the relationship between
the space locations of pixels and make it evenly
distributed in all space of the carrier image. This will
improve the robustness of the algorithm. Two-
dimensional Arnold scrambling transformation is
defined as follows:
X′
Y′
11
12
X
Y
modN
(1)
wherein, X and Y are the pixel coordinates of the
original space: X', Y' are the corresponding pixel
coordinates after iterative computation scrambling.
The parameter N represents the size of the
rectangular image, also referred to as a step number.
To restore the original initial watermark, the
corresponding inverse transform formula is applied,
it is given by:
X′
Y′
2 1
1 1
X
Y
modN
(2)
Arnold transformation is cyclical. If the cycle
and the number of iterations are not known, the
watermark is not restored. Therefore, cycle and
iterations can exist as a private key.
2.2 Embedding Process
2.2.1 Fingerprint Pre-Processing
This task is performed by removing the noise using
region based segmentation method, which presumes
that pixels in the same regions are similar in such
brightness and texture. After that the Region of
Interest (ROI) is extracted and cropped from the
denoising image. In proposed system, the fingerprint
center point is automatically detected using the
method described in (
Julasayvake and Choomchuay,
2010) for the cropping region. Depending on this
center point, the image of size N
N is cropped and
used for watermarking process.
2.2.2 Embedding Process in One Level
The embedding process is applied in the spatial
domain by using the original LBP operator and
Boolean function operations defined in (Shih and al.,
2011). The insertion is performed by adjusting one
or more of the pixels in the neighborhood to make
the Boolean function results consistent with the bits
of the watermark. In our scheme, the watermark is
composed of ID image and grayscale face image.
The schematic diagram of the proposed embedding
process is given in figure 3.
Step 1: The original fingerprint image I(i,j) is
subdivided in G non-overlapping blocks of 3x3, to
which the LBP operator (
Ahonen et al., 2006) is
applied to calculate the magnitude matrix
and
the matrix sign
. The matrix
is constructed by
calculating the absolute values of the difference
between the gray level of the center pixel
and its
8 neighborhoods
as follows:
/
|
|,0,….,7
(3)
The matrix
is obtained by applying the operator
LBP of the matrix G, it is given by:
S
S
/S
sgng
g
,i0,….,7
(4)
where sgn refers to the sign function defined as:
f
x
1, 0
0, others
(5)
Step 2: Calculating the Boolean function
by
applying the XOR operator (⊕) on the binary sign
vector
as follows:
⊕
⊕…⊕
(6)
ProtectingDigitalFingerprintinAutomatedFingerprintIdentificationSystemusingLocalBinaryPatternOperator
35
Figure 3: Schematic diagram of proposed embedding
process.
Step 3: performs the embedding operation: to embed
one bit of the watermark, we search the embedding
location by comparing the value of the
function
with the bit value w
k
of the
watermark. If they are different, we search the
location I
,
of the minimum value
,
in the
matrix
and, we modify the pixel corresponding
to the similar position in the matrix G
p
. Otherwise,
we do nothing to the pixels in the neighborhood. The
insertion operation is as follows:
if (w
k
!= f (S
p
)) then
{
select m
i,j
=min (M
p
)
if (S
i
,
j
==1) then
g'
i
,
j
= g
i
,
j
+ [-ßg
i
,
j
+m
i,j
*(ß-
1)]*
,
;
else g'
i
,
j
= g
i
,
j
+ [ßg
i
,
j
+m
i,j
*(ß+1)]*
,
}
The parameter ß represents the strength factor.
takes the value0 if the pixel (i,j) under consideration
belongs to a fingerprint feature region like delta or
core areas (singular points); it has value 1 otherwise.
To achieve higher embedding payload and better
robustness, the embedding process is extended to
double level. By this, we can insert not only the ID
image of the person, but also other information such
as the image of his face.
2.2.3 Embedding Process in Double Level
To perform the embedding in double level, the
neighborhood
calculated in aforementioned
embedding operation is divided in two parts even
and odd neighbors, denoted by
and
respectively(Figure 4). After that, the
functions
and
are computed and
according to their values, two bits of the watermark
are hidden such that each part hides one bit of the
watermark. The embedding is done conforming to
insertion process given above. In this way, the
watermarking capacity is doubled.
g
3
g
2
g
1
g
odd
={g
0
, g
2
, g
4
, g
6
}
g
even
={g
1
, g
3
, g
5
, g
7
}
g
4
g
c
g
0
g
5
g
6
g
7
Figure 4: Partition of the matrixS
in two sets evenS
and oddS
.
2.3 Watermark Extraction Process
The detection is performed in a blind manner which
means that the original fingerprint image is not
needed at the extraction. The watermark bits W
k
are
extracted from each part (odd and even) within a
block after having calculated the matrix
and
. According to their values,
extracted bits are determined as follows:
Yes
Start
Fingerprint image & Watremarks
Subdivide the in
p
ut ima
g
e in N blocks of size 3*3
Yes
p=1
if f(S
p
)
w
p
No
No
Search the min
,
in the matrix M
p
End
Watremarked Image
if
p=p+1
Embed w
p
in
g
i,j
Fingerprint preprocessing
Calculation of the Sign matrix (S
p
)
and Ma
g
nitude matrix
(
M
p
)
Calculation f(S
p
)
SIGMAP2014-InternationalConferenceonSignalProcessingandMultimediaApplications
36
(7)
3 EXPERIMENTS AND RESULTS
To gauge the performance of the proposed scheme,
we carried out the simulation on fingerprint images
of size (248*292). The watermark is composed of a
binary ID image of size 240x22 and grayscale face
image of size 35x35. This conducts to the embedded
watermark data of (16092 bits).
The imperceptibility property determines how
much the watermarked fingerprint image differs
from the original fingerprint image; in other words,
how much the embedding process distorts the host
image. In this paper, the conventional image quality
metric Peak to Signal-to-Noise Ratio (PSNR) is used
as the criteria of imperceptibility. It is computed as
follows:
10.
255
(8)
with
∑∑
,
,
∗
(9)
where f(x,y) and f’(x,y) represent the pixel values of
the original host fingerprint image and the
watermarked image respectively. The parameters x,
y specify row and column size of images
respectively.
The extraction performance of watermarking is
measured by the Error Bit Rate (EBR) or the error
probability. The EBR is the number of extracted bits
that have been altered due to noise, interference and
distortion, divided by the total number of embedded
bits. EBR is a dimensionless performance measure,
often expressed as a percentage number (
Xinhong, et
al., 2012)
. It is given by:
∑∑
,
⊕
,
∗
(10)
where
,
,and
, denote the original
watermark and the extracted one of (n*m) size
respectively.
As it can be seen from figure 5, the watermarked
fingerprint images is decoded with 100 % decoding
accuracy; also the watermarking does not change the
fingerprint features of the original images as well the
visual quality. This results in the PSNR equal to 36
dB.
In addition to imperceptibility and capacity
criteria, the reliability of the watermarking approach
correlates also with the robustness against the
attacks. The main requirement of robustness is to
resist different kinds of distortions introduced by
common processing or malicious attacks while
satisfying the imperceptibility criteria. Table 1
shows the robustness of the proposed method against
some image-processing manipulations such as
luminance manipulation, contrast enhancement,
additive noise and JPEG compression.
Figure 6 shows the EBR curves after applying above
mentioned manipulations using the two modes of
insertion: simple embedding in which one bit of the
watermark is embedded within the block of size 3x3
and double level of embedding where the block is
partitioned in two parts even an odd. We notice that
in double-level watermarking process the robustness
against manipulations is better than in the simple
embedding one despite the payload is two times
larger.
The proposed scheme provides excellent results
in term of the payload compared to the results
obtained in (Jain and Uludag,
2003)
and (Pravin et
al., 2011) while keeping the subjective visual quality
unchanged.
Figure 5:Subjective visual quality: (a) Original fingerprint
image, (b), watermarked fingerprint image.
In term of protecting the biometric features of
fingerprint image, in our scheme, we propose to
protect the fingerprint feature region like delta or
core areas by masking their pixels. The pixel takes
the value 0 if the pixel under consideration belongs
to a fingerprint feature otherwise it takes 1. Contrary
to (Pravin et al., 2011) where the authors did not
discuss how the biometric features are protected.
ProtectingDigitalFingerprintinAutomatedFingerprintIdentificationSystemusingLocalBinaryPatternOperator
37
Table 1: Extracted watermarks resulting after some
applying spatial attacks.
Attacked
watermarked
image
Extracted
Watermarks
EBR
(%)
Cropping
7.85
2.70
Additive noise
1.69
1.77
Luminance
manipulation
0.46
0.8
Contrast
enhancement
0.2
0.5
JPEG
Compression with
quality 100
5.5
4.32
4 CONCLUSIONS
In this paper, a local adaptive watermarking method
based on LBP operator is presented to provide data
integrity and authenticity of the fingerprint images
content. The embedding consists of hiding the
identifier (ID) image and the face image in the
corresponding fingerprint image after being enrolled
by the AFIS system. This process eliminates data
mismatch, reduces the high cost of storage, speeds
the retrieval of related data and detects tampering.
The results show that the criterion for
imperceptibility is achieved; the fingerprint images
are watermarked without changing the features
associated with them. The robustness against some
spatial attacks is provided. However the proposed
scheme is not prove to be robust to WSQ (Wavelet
Scalar Quantization) compression.
As fingerprint images are often compressed using
an open wavelet-based image compression (WSQ)
developed by the FBI, we are working to improve
the robustness against WSQ by introducing the error
code coding (ECC) before the embedding process
and doing the embedding in frequency domain.
Figure 6: Resulting EBR against the spatial attacks.
EBR
EBR
EBR
EBR
(a)Noise
(d) Contrast adjustment
(c) Luminance
(b) JPEG compression
One-level One level watermarking
Double-level watermarking
SIGMAP2014-InternationalConferenceonSignalProcessingandMultimediaApplications
38
REFERENCES
Ahonen, T., Hadid, A., Pietikäinen, M., 2006. Face
description with local binary patterns: Application to
face recognition. IEEE Transactions on Pattern
Analysis and Machine Intelligence, pp. 2037-2041.
Jain, A. K., Uludag,U.,2003. Hiding biometric data. IEEE
Trans. Pattern Anal. Machine Intell., vol. 25, no. 11,
pp. 1494–1498.
Julasayvake, A., Choomchuay, S., 2010. An algorithm for
fingerprint core point detection. International Journal
of Computer Applications, vol. 2, n°. 8, pp. 0779 –
4244.
Mathivadhani, D., Meena, C., 2010. A comparative A
Comparative Study on Fingerprint Protection Using
Watermarking Techniques. Global Journal of
Computer Science and Technology, vol. 9, Issue 5, pp.
98-102.
Noore, A., Singh, R. Vatsa, M. and Houck, Max M., 2007.
Enhancing Security of Fingerprints through
Contextual Biometric Watermarking. Forensic Science
International, Issue 169, no. 2, pp. 188-194.
Uludag, U., Jain, K., 2004. Attacks on Biometric Systems:
A Case Study in Fingerprints. In.Proc. SPIE-EI,
Security, Seganography and Watermarking of
Multimedia Contents VI.
Wenyin, Z., Shih, F.Y., Faugeras, 2011. Semi-fragile
spatial watermarking based on local binary pattern
operators. Elsevier, 3904–3912.
Xinhong, Z. Fan, Z., LI. D and Li, C., 2012. An Analysis
of Relationship of Watermarking Decoding Error Bit
Rate and Payload Capacity. Journal of Computational
Information Systems vol. 8, n°. 20, pp. 8431–8438
Available at http://www.Jofcis.com
yothish. J., Prabhu, V., Kumar, S., 2012. A Robust
Watermarking method based on Compressed Sensing
and Arnold scrambling. IEEE Machine Vision and
Image Processing (MVIP), pp.105-108.
ProtectingDigitalFingerprintinAutomatedFingerprintIdentificationSystemusingLocalBinaryPatternOperator
39
