Secure_and_robust_user_authentication_using_partial_fingerprint_matching
Secure_and_robust_user_authentication_using_partial_fingerprint_matching
Geuntae Bae, Hojae Lee, Sunghoon Son, Doha Hwang, Jongseok Kim
Samsung Electronics, Korea
{geuntae.bae, hojae72.lee, s.h.son, doha.hwang, jx.kim}@samsung.com
Authorized licensed use limited to: Indian Institute of Information Technology Kottayam. Downloaded on February 04,2025 at 12:39:54 UTC from IEEE Xplore. Restrictions apply.
978-1-5386-3025-9/18/$31.00 ©2018 IEEE
The non-minutiae based approaches often utilize additional II. BACKGROUND AND RELATED WORK
characteristics from the fingerprint ridge pattern, such as local
orientation and frequency, ridge shape, and texture information A. Fingerprint Image Enhancement
containing higher discriminatory power [8]. Among them, the A variety of image enhancement methods has been
correlation-based approach superimposes two fingerprint proposed and being used, as the fingerprint pre-processing [1]
images and calculates the correlation between corresponding [2] [18] [19]. In this work, we used techniques such as
pixels in varied alignments, to estimate the similarity of the “Normalization” to stretch the contrast of an image, “ROI
two fingerprints [10] [11]. Segmentation” to locate the effective area of a fingerprint,
“Band-pass filter” to strengthen the ridge frequency signal, and
To accelerate the alignment or matching, some algorithms “Ridge-orientation filter” to boost up the signal in the ridge
use “Phase-only correlation (POC)” technique [12], and some direction.
use “Fourier-Mellin transform (FMT)” additionally for rotation
and scale matching [13] [14]. The POC has been successfully
applied to high-accuracy image registration problems in
computer vision application. The advantages of the POC are its
numerical efficiency, and robustness against noise by its high
resolution.
While most recent algorithms employ hybrid approach that
properly combine these various methods [15] [16] [17],
basically the correlation-based method is suitable for partial
fingerprints of good quality, by its high discriminatory
performance.
In designing a biometric verifier, performance factor such
as recognition rate, is obviously a top consideration since it is
directly related to usability [18] [19]. In addition, security
aspects such as defensiveness against explicit attacks should
also be considered. It is well known that in spite of its various
advantages, fingerprint-based biometric systems are potentially
vulnerable to a variety of attacks [20] [21] [22]. In particular,
poor information entropy of the partial fingerprint area should Figure 2 Fingerprint Enhancement
not be overlooked.
B. Minutiae Extraction
In sum, the problem description of this work is as follows:
Locating Minutiae points in the fingerprint image is also
• Mobile phones and other recent small form factor one of the well-researched subjects in fingerprint processing [5]
devices generally prefer acquiring only a partial portion [6] [7].
of the full fingerprint. Partial fingerprints possibly lack
For this work, processing such as “Ridge-direction filter”,
a sufficient number of feature patterns called ‘minutiae’,
“Binarization”, “Thinning (Zhang Suen)”, and “Elimination of
necessarily used by the most common method of the
false minutiae” [26] were used in a proper combination.
fingerprint matching.
• Good usability requires multiple enrollment of partial
fingerprints from a single finger to accommodate the
variation of finger impress position. In cases, a single
identity or person is allowed to enroll multiple fingers.
Hence, commonly a verifier declares accept if an input
fingerprint match at least any one of a large number of
the stored templates, thus it might eventually result in
security risks.
• For enhanced biometric performance and usability, fast
and precise processing even with the image variability
by finger rotation, position, deformation, etc. is critical.
The rest of the paper is organized as follows. In section 2,
we study the related work regarding the fingerprint image
processing algorithms. Section 3 first outlines a design goal of
the target system in security and performance aspects. Then,
we discuss details of the implemented algorithm and reasoning.
The result of experimental evaluations on our work is
presented in Section 4. Figure 3 Minutiae (feature) Extraction
Authorized licensed use limited to: Indian Institute of Information Technology Kottayam. Downloaded on February 04,2025 at 12:39:54 UTC from IEEE Xplore. Restrictions apply.
C. FFT-based Image Registration ݄ሺݔǡ ݕሻ ൌ ݃ሺߠ ݔ ߠ ݕ െ ܽǡ
Image Registration is also well-researched area in computer െߠ ݔ ߠ ݕ െ ܾሻ
image processing [23][24][25]. Phase correlation provides According to the Fourier translation and rotation property,
estimation of rigid translational displacement between two transform of ݃ሺݔǡ ݕሻ, ݄ሺݔǡ ݕሻ are now related by
images, which is based on the well-known Fourier shift
theorem: a shift in the spatial domain of two images results in a ܪሺݑǡ ݒሻ ൌ ܩሺߠ ݑ ߠ ݒ ǡ െߠ ݑ ߠ ݒ ሻ
linear phase difference in the frequency domain by the Fourier ൈ ݁ ିଶగሺ௨ା௩ሻ
Transform (FT).
The magnitudes of ܩሺݑǡ ݒሻ and ܪሺݑǡ ݒሻ only represent a
Given two functions ݃ሺݔǡ ݕሻ, ݄ሺݔǡ ݕሻ representing two 2D relative angle(ߠ ሻ, excluding the translation, as follows.
images different only in a translational shift (ܽ in horizontal,
ܪ ሺݑǡ ݒሻ ൌ ܩ ሺߠ ݑ ߠ ݒ ǡ െߠ ݑ ߠ ݒ ሻ
and ܾ in vertical directions),
Then if we consider ܩ ሺݑǡ ݒሻ and ܪ ሺݑǡ ݒሻ in polar
݄ሺݔǡ ݕሻ ൌ ݃ሺ ݔെ ܽǡ ݕെ ܾሻ
representation, ܩ ሺߠǡ ߩሻ ൌ ܪ ሺߠ െ ߠ ǡ ߩሻ
The corresponding Fourier Transforms are denoted by
Much less computation is sufficient for the rotation
ܩሺݑǡ ݒሻ and ܪሺݑǡ ݒሻ.
matching that searches only the rotation angle without having
to consider the scale change. That is, since only Ʌ of the polar
ܪሺݑǡ ݒሻ ൌ ඵ ݄ሺݔǡ ݕሻ݁ ିଶగሺ௫௨ା௬௩ሻ ݀ ݕ݀ݔሺ ሻ domainሺɅǡ ɏሻ is to be found through phase correlation, only 1-
dimensional iFT of Ȫ൫ሺሻ൯ is required as follows.
ൌ ඵ ݃ሺ ݔെ ܽǡ ݕെ ܾሻ݁ ିଶగሺ௫௨ା௬௩ሻ ݀ ݕ݀ݔ
Authorized licensed use limited to: Indian Institute of Information Technology Kottayam. Downloaded on February 04,2025 at 12:39:54 UTC from IEEE Xplore. Restrictions apply.
Figure 7 Verification Procedure (algorithm)
matching. Since the Fourier magnitude spectrum contains a lot images(64x64), the calculations can be further lowered in
of information about the image area, a strong correlation at a comparison with the whole image(128x128).
certain relative angle implies that the two image areas are
likely to be actually very similar. In addition, the band-limited POC can be used [12] [16].
Filtering by a POC peak threshold can be also used for
However, experimental observations have stated that the efficiency, since low peak magnitudes strongly imply low
rotational correlation for whole area cannot guarantee high similarities of the images in comparison.
accuracy, since the non-overlapping area act as noise. By
dividing an image into segments and searching the max value 3) Scoring (MS): Feature-weighted block scoring
of the phase correlation peaks for all segment combinations, As the last step, scoring or similarity evaluation is
the accuracy can be greatly improved. conducted. Basically, the similarity of two superimposed
images can be calculated by “Normalized Cross Correlation
(NCC)”, as follows. (0 < NCC < 1)
σ௫ǡ௬൫݃ሺ ݔെ ܽǡ ݕെ ܾሻ ݄ ڄሺݔǡ ݕሻ൯
ሺ ሻ ൌ
ඥσ௫ǡ௬ ݃ሺ ݔെ ܽǡ ݕെ ܾሻଶ ڄσ௫ǡ௬ ݄ሺݔǡ ݕሻଶ
Thus in a correlation-based method, the Match Score (MS)
can be quantified using the NCC, the size of the overlapped
area of two images, and so on.
For accuracy and efficiency, the similarity evaluation is
conducted per block based and cumulative-based, as follows.
Authorized licensed use limited to: Indian Institute of Information Technology Kottayam. Downloaded on February 04,2025 at 12:39:54 UTC from IEEE Xplore. Restrictions apply.
In addition, by numerous experimental observations on IV. EVALUATION
partial fingerprint images, we found out that minutiae areas are
much more discriminatory than other regions of an image. A. Experimental Environments
Thus it is reasonable to assign a greater weight to a minutiae Capacitive, 508 dpi,
block than to other blocks, in calculating scores. Fingerprint sensor Acquisition size : width:160p, height:70p
(physical effective area: 8.0 x 3.5mm)
100 persons (3 finger per person)
- 25 enroll images per finger
Fingerprint database
- 180 query images per finger
(data acquisitions)
(60 query images for 3 variant degree
acquisition)
Test platform (board) Samsung Exynos 8890
B. Performance Evaluation
Verification tests were conducted on android platform with
database acquired by the capacitive fingerprint sensor for the
Thus, we have created an enhanced calculation method for latest mobile device. The database was obtained directly from
scoring that reflects this consideration. 100 volunteers and the number of test cases is considered
sufficient for the biometric evaluation.
- Total Num. of imposter verifications: 5,346,000 ([[[[)
- Total Num. of genuine verifications: 54,000 ([[[[)
Authorized licensed use limited to: Indian Institute of Information Technology Kottayam. Downloaded on February 04,2025 at 12:39:54 UTC from IEEE Xplore. Restrictions apply.
[7] Jea, Tsai-Yang, and Venu Govindaraju. "A minutia-based partial
fingerprint recognition system." Pattern Recognition 38.10 (2005):
1672-1684.
[8] Yang, Jucheng. Non-minutiae based fingerprint descriptor. INTECH
Open Access Publisher, 2011.
[9] A. K. Jain, Y. Chen and M. Demirkus, "Pores and Ridges: High-
Resolution Fingerprint Matching Using Level 3 Features," in IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 29, no.
1, pp. 15-27, Jan. 2007.
[10] Karna, Deepak Kumar, Suneeta Agarwal, and Shankar Nikam.
"Normalized cross-correlation based fingerprint matching." Computer
Graphics, Imaging and Visualisation, 2008. CGIV'08. Fifth International
Conference on. IEEE, 2008.
[11] Cavusoglu, Abdullah, and Salih Gorgunoglu. "A robust correlation
based fingerprint matching algorithm for verification." Journal of
Applied Science 7 (2007): 3286-3291.
[12] Koichi, I. T. O., et al. "A fingerprint matching algorithm using phase-
only correlation." IEICE Transactions on Fundamentals of Electronics,
Communications and Computer Sciences 87.3 (2004): 682-691.
[13] J. Zhang, Z. Ou and H. Wei, "Fingerprint Matching Using Phase-Only
Correlation and Fourier-Mellin Transforms," Sixth International
Conference on Intelligent Systems Design and Applications, Jinan,
2006, pp. 379-383.
[14] Jin, Andrew Teoh Beng, David Ngo Chek Ling, and Ong Thian Song.
"An efficient fingerprint verification system using integrated wavelet
and Fourier–Mellin invariant transform." Image and Vision Computing
Evaluation with top 5% common pattern imposters 22.6 (2004): 503-513.
[15] Ross, Arun, Anil Jain, and James Reisman. "A hybrid fingerprint
The verification test by imposters gathered from the false matcher." Pattern Recognition 36.7 (2003): 1661-1673.
match score record of upper 5% which contain primarily
[16] Ito, Koichi, et al. "A fingerprint recognition algorithm combining phase-
common pattern and rarely include minutiae, has shown a little based image matching and feature-based matching." International
degraded result. But it remained at a reasonable level (3.1%), Conference on Biometrics. Springer Berlin Heidelberg, 2006.
thus proved to be considerably robust also in security aspect. [17] Nandakumar, Karthik, and Anil K. Jain. "Local Correlation-based
Fingerprint Matching." ICVGIP. 2004.
V. CONCLUSION [18] KHAN, Muhammad Ali., Fingerprint image enhancement and minutiae
In this work, we presented a novel partial fingerprint-based extraction. 2011.
verifier that ensures secure and robust user verification [19] Lin Hong, Yifei Wan and A. Jain, Fingerprint image enhancement:
algorithm and performance evaluation, in IEEE Transactions on Pattern
especially for mobile environment and small-sized sensors. Analysis and Machine Intelligence, vol. 20, no. 8, pp. 777-789, Aug
The method of "segmented area matching" strongly 1998.
enhanced robustness to the query variability, especially to [20] Roy, Aditi, Nasir Memon, and Arun Ross. "MasterPrint: Exploring the
Vulnerability of Partial Fingerprint-based Authentication Systems."
finger rotation. And the method of "feature-weighted block IEEE Transactions on Information Forensics and Security (2017).
correlation scoring" provided much more detailed image
[21] Jea, Tsai-Yang, et al. "Security and matching of partial fingerprint
discrimination, resulting in improved security. Experimental recognition systems." Defense and Security. International Society for
evaluations with extensive database of partial fingerprint Optics and Photonics, 2004.
images from more than 100 people, acquired by a small-sized [22] U. Uludag and A. K. Jain. Attacks on biometric systems: a case study in
capacitive sensor, demonstrated a significant improvement fingerprints. In SPIE Electronic Imaging, pages 622–633. Intl Society
over the traditional correlation-based methods. for Optics and Photonics, 2004.
[23] Zitova, Barbara, and Jan Flusser. "Image registration methods: a
REFERENCES survey." Image and vision computing 21.11 (2003): 977-1000.
[1] Maltoni, Davide, et al. Handbook of fingerprint recognition. Springer [24] Reddy, B. Srinivasa, and Biswanath N. Chatterji. "An FFT-based
Science & Business Media, 2009. technique for translation, rotation, and scale-invariant image
registration." IEEE transactions on image processing 5.8 (1996): 1266-
[2] Ratha, Nalini, and Ruud Bolle, eds. Automatic fingerprint recognition 1271.
systems. Springer Science & Business Media, 2007.
[25] Tzimiropoulos, Georgios, et al. "Robust FFT-based scale-invariant
[3] Zhang, David, et al. "Selecting a reference high resolution for fingerprint image registration with image gradients." IEEE transactions on pattern
recognition using minutiae and pores." IEEE Transactions on analysis and machine intelligence 32.10 (2010): 1899-1906.
Instrumentation and Measurement 60.3 (2011): 863-871.
[26] Kim, Seonjoo, Dongjae Lee, and Jaihie Kim. "Algorithm for detection
[4] Jain, Anil K., Yi Chen, and Meltem Demirkus. "Pores and ridges: High- and elimination of false minutiae in fingerprint images." International
resolution fingerprint matching using level 3 features." IEEE Conference on Audio-and Video-Based Biometric Person
Transactions on Pattern Analysis and Machine Intelligence 29.1 (2007). Authentication. Springer Berlin Heidelberg, 2001.
[5] Zaeri, Naser. "Minutiae-based fingerprint extraction and recognition", [27] Mathur, Surbhi, et al. "Methodology for partial fingerprint enrollment
INTECH Open Access Publisher, 2011. and authentication on mobile devices." Biometrics (ICB), 2016
[6] Xiping Luo, Jie Tian and Yan Wu, "A minutiae matching algorithm in International Conference on. IEEE, 2016.
fingerprint verification," Proceedings 15th International Conference on
Pattern Recognition. ICPR-2000, Barcelona, 2000, pp. 833-836 vol.4.
Authorized licensed use limited to: Indian Institute of Information Technology Kottayam. Downloaded on February 04,2025 at 12:39:54 UTC from IEEE Xplore. Restrictions apply.