Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: Today’s with increasing identity theft, biometric systems based on fingerprints have a growing impor-
Received 7 January 2019 tance in protection and access restrictions. Malicious users violate them by presenting fabricated
Revised 30 April 2019 attempts. For example, artificial fingerprints constructed by gelatin, Play-Doh and Silicone molds may
Accepted 16 June 2019
be misused for access and identity fraud by forgers to clone fingerprints. This process is called spoofing.
Available online 4 July 2019
To detect such forgeries, some existing methods using handcrafted descriptors have been implemented
for assuring user presence. Most of them give low accuracy rates in recognition. The proposed method
Keywords:
used Discriminative Restricted Boltzmann Machines to recognize fingerprints accurately against fabri-
Biometric systems
Deep learning
cated materials used for spoofing.
Discriminative Restricted Boltzmann Ó 2019 Karabuk University. Publishing services by Elsevier B.V. This is an open access article under the CC
Machines BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Fingerprint authentication
https://doi.org/10.1016/j.jestch.2019.06.005
2215-0986/Ó 2019 Karabuk University. Publishing services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
D.M. Uliyan et al. / Engineering Science and Technology, an International Journal 23 (2020) 264–273 265
which can be trained to distinguish a live finger from a spoof shown in Fig. 2 arrange two types of layers: a visible layer v with
[6,43]. Deep learning has been adopted to generate a robust finger- stochastic m neurons are direct relations with a hidden layer h
print spoofing system. with n neurons. However, there are no connections between neu-
For instance [31], shows that convolution neural network can rons in the same layer, and this restriction gives the RBM its name.
identify the forged fingerprints with known fabricate materials. A weight matrix is constructed asWmXn . It has the weights between
Chugh et al. [44] proposed a method for fingerprint recognition the connected, visible and hidden neurons, where wij represents
by extracting local regions from the image itself. These patches symmetric weights between the ith visible neuron vi connected
were centered and aligned using minutiae features to provide sali- with the jth hidden neuron hj .
ent descriptors. These cues are trained via CNN models to improve Each unit in a visible layer V is connected with all the units in
detection results. the hidden layer h. Assume layers, V and h, with binary stochastic
In this paper, we proposed a deep learning method for to recog- variables, i.e., V 2 ½0; 1m and h 2 ½0; 1n . Where m and n are the
nize real fingerprints and detect spoofs. The main contributions of numbers of units of the visible and hidden layers. The energy func-
our method are summarized as follows: tion for connecting configuration between V and h layer from Ber-
noulli distribution with success probability in RBM method [48] is
1) It attempts to distinguish real fingerprint images from fake defined as
ones.
X
m X
n X
m X
n
2) It analyzes the image consistency based on scaled and EðV; hÞ ¼ ai vi b j hj vi hj wij ð1Þ
rotated ROIs. Based on these ROIs features, we propose a i¼1 j¼1 i¼1 j¼1
deep discriminative model for training detection.
where ai and bj are the biases of ith visible layer and jth hidden
The rest of the paper is presented as follows. In Section 3, we layer, respectively. The model parameters ai is composed of vector
introduce two main techniques: DRBM and DBM used in the pro- a ¼ ½a1 ; a2 ; :; aV and bj is composed of vector b ¼ ½b1 ; b2 ; :; bh .
posed method identify representation of real and forged finger- The marginal probability of a joint distribution function over visible
prints. Experimental results are presented in Section 4. Finally, layer is defined as
Section 5 gives a conclusion and future works. 1X
PðvÞ ¼ expEðv;hÞ ð2Þ
Z h
3. Proposed method
where Z is the partition function defined as follows:
XX
We propose a novel method to determine fake fingerprints Z¼ expEðv;hÞ ð3Þ
regarding spoof forgeries were used in authentication based sys- v h
tems, we adapt multiple features called deep features which
Z is the sum of possible pairs of (v, h). Let V be a m dimensional vec-
extracted from images such as, Deep Boltzmann Machine (DBM)
tor and let h be a n dimensional binary vector. while visible units
[45]. The main benefit of DMB is that its layered architecture helps
are binary, we have total 2mþn pairs of ðv; hÞ.
to investigate a complex relationships between features and
Since the RBM is a probabilistic binary structure may includes
enables deep learning of high detailed features of data. Extracting
different layers, each higher layer takes the correlation between
features from the input image based on Deep Neural Networks aids
actions of upper hidden features from lower hidden layer. The
to understand data in depth. DBMs are probabilistic deep learning
lower layer generates visible units connected with the higher layer
that copes with complex patterns impressively by extracting
based on joint distribution with mutual independent conditional
highly detailed features from the image. It is suitable for tasks in
distribution probabilities PðhjVÞ and PðVjhÞ are defined as follows
which the patterns might not be easily detected or forged. Results
on the Cross Match [7] dataset show that our technique achieves X
n
good results with comparison to handcrafted based methods, the Pðvi ¼ 1jhÞ ¼ /ð wij hj þ ai Þ ð4Þ
j¼1
results show high detection rate regarding spoofing attack. The
motivation of using deep learning features includes: and
X
m
1) Restricted Boltzmann Machines (RBMs) adapted in many Pðhj ¼ 1jvÞ ¼ /ð wij vi þ bj Þ ð5Þ
applications such as image classification and medical image i¼1
investigation and pattern analysis to solve various learning
problems. where / () stands for the sigmoid function, /ðxÞ ¼ 1=ð1 þ expðxÞÞ.
2) Restricted Boltzmann Machines typically developed to For an nhidden-layer RBM, its model parameters consist of
extract features and build a self-contained framework for value
generating competitive non-linear classifiers. T ¼ fa1 ; b1 ; w1 ; a1 ; b1 ; w1 ; :; am1 ; bn1 ; wmn1 ; am ; bn ; wmn g
3) We introduce RBM algorithm that introduces a discriminant
ð6Þ
factor to RBM training.
Given the training input data T of the visible neuron, it is not
Thus, the notion of two main deep learning features: Restricted easy to approximate the parameters of vector T by using maximum
Boltzmann Machines (RBM) and Deep Boltzmann Machines (DBM) likelihood criterion directly because of various hidden layers found
are introduced below regarding spoofing forgeries. in the RBM model. To solve this issue, we have implanted a greedy
based learning algorithm to train the stacked RBM model [49] in
the input image. A greedy learning algorithm tries to find out the
3.1. Restricted Boltzmann Machines (RBM)
parameters of the first layer RBM a1 ; b1 ; w1 to model the visible
The (RBM) as introduced in [46] is a kind probabilistic graphical training data. Then, it saved the parameters of the first layer into
model based stochastic neural networks which are used to vector v ¼ a1 ; w1 and produce a Gibbs sampling from the first
describe the dependency among a set of random data variables layer in RBM as Pðhj ¼ 1jvÞ to train the next layer of RBM
using a two layer architecture. The architecture of an RBM as a2 ; b2 ; w2 , Let h = (a,b,W) be the set of parameters of RBM. The
D.M. Uliyan et al. / Engineering Science and Technology, an International Journal 23 (2020) 264–273 267
Fig. 3. a) three-layer Deep Boltzmann Machine. b): Pre training consists of learning a stack of modified RBM’s, which are then composed to create c) a deep Boltzmann
machine.
268 D.M. Uliyan et al. / Engineering Science and Technology, an International Journal 23 (2020) 264–273
4. Experiment results the influence of modifying learning rate and weight decay param-
eters in the proposed method.
4.1. Datasets
4.4. Self comparison
In this section, we used the state of art LivDet 2datasets: LivDet
2013 [53], and LivDet 2015 for fingerprints spoofing benchmark DRBM + DBM method is trained on three data sets for finger-
[20,25] which are considered in the proposed method. The LivDet print detection. We use two groups of patches generated from
2013 Contains two type of images: genuine images and spoofs with the input image, Scaled and rotated ROIs. To evaluate progress of
total of 16.000 images and equally distributed as 50% of the images our method, three fingerprint scanners for spoof images were
were used as a training set and the remaining 50% as the test set for regarded in our experiments. We have used the following metrics
classification. as shown in Table 3: Accuracy rate (ACC), Half total error rate
Hence, the real images captured by four fingerprint scanners: (HTER), Fake Fingerprint Accuracy (FFA) and True Fingerprint
Biometrika, Crossmatch, ItalData and Swipe, Spoofs are created Accuracy (TFA), respectively and defined as follows
using a gelatin materials like latex, play-doh and wood glue. LivDet
NT
2013 utilized for use of the non-cooperative method without user ACC ¼ 100% ð12Þ
interference for creating spoof images. We used only Biometrika
NT þ NF
and ItalData due to high resolution of their images. The second where NT represents the number of the correct fingerprint recog-
dataset is livDet 2015 which contains a set of images were cap- nized in the testing dataset and NF is the number of fake finger-
tured by four optical scanners; these scanners are Digital Persona, prints in testing dataset.
Green Bit, Biometrika and Crossmatch. It noticed that the testing
FAR þ FRR
set has spoof images which forged using unknown materials not HTER ¼ ð13Þ
saved in the training set. Only gelatin material in the Crossmatch 2
were considered in our experiments as shown in Table 1. where FAR is a False acceptance rate and FRR is a false rejection
Artificial fingerprints in Biometrika and Italdata have been cre- rate. In order to assess the performance of the proposed method
ated without user assistance, while fake ones in CrossMatch were based on various materials for spoofing forgeries, fake fingerprint
created by user cooperation. Several materials for creating the syn- accuracy (FFA) and true fingerprint accuracy (TFA) are defined as
thetic fingerprints were applied, such as: gelatin, wood glue, sili- follows:
cone and latex. More concisely, experimental evaluation of the
N1
proposed method is applied on a collection of a large and challeng- FFA ¼ 100% ð14Þ
ing LivDet 2013 and 2015 dataset. In LivDet 2013, over 1000 live NF
attempts were collected as well as, 1000 spoof fingers from various
N2
materials such as Play-Doh, Gelatin and Ecoflex. In LiveDet 2015, it TFA ¼ 100% ð15Þ
NT
is about 2011 training attempts have conducted on 1010 live fin-
gers with a resolution of 1000 dpi from 51 subjects depending on where N1 represents the number of fake fingerprint images which
the sensor, 2 fingerprints each of all 10 fingers and 1001 spoofed are recognized falsely in NF. N2 represents the number of true fin-
fingerprints across 5 spoof materials to produce nearly 200 fin- gerprint images that, are identified as true ones in NT.
gerprint images per spoof material. 500 spoofed fingers made from As pointed in Table 3, the computation cost of the proposed
each of 5 fingerprints of 20 subjects for each of the five spoof mate- method for a single tested image is 45 s when we test images on
rials. Only two attempts applied on each spoof. Italdata dataset. Each step is given individually. Fingerprint prepro-
cessing step for one image required 0.2 s and fingerprint deep fea-
tures extraction required 10 s. The DBM training consumed 30 s
4.2. Experimental setup and KNN classifier required 4.8 s. Similarly, the total computation
time required to test one image on CrossMatch is 85 s, which is
To achieve a perception of these three fingerprint benchmarks, slower than Italdata and Biometrica due the large blank areas.
various parameters in our method have been adjusted to give a We have tested three types of spoof materials: Wood Glue, gelatin,
good results as shown in Table 2. The implementation of the pro- Play-Doh and 2D-printed ones on the proposed method.
posed method was applied on Intel (R) Xeon (R) CPU E5-2690 v2 To examine the time complexity of the proposed method, two
(3.00 GHz processor) with 20 GB RAM and NVIDIA GPU uses Mat- experiments are conducted. The experiments considered the num-
lab 2016a. ber of training fingerprints as well as number of input features
used in the tested fingerprint image. The computation times of
the proposed DRBM + DBM model are gained for different numbers
4.3. Parameters evaluation of training images. The LivDet dataset is employed in the first
experiment. The number of training images is ranged from 50 to
To do training and testing samples of fake fingerprints in the 1000. The average processing time of the proposed method is cal-
dataset, we have tried various values of specific parameters: patch culated against the number of training fingerprint images as shown
numbers, learning rate and weight decay respectively on our in Fig. 10. It shows that the time complexity of the proposed
experiments. Appropriate values for these parameters are chosen. method in regard to the number of training samples is O (n) based
As shown in Fig. 7, Using the 10 ROIs from the original input finger- on big O (.) notation.
print in the training DBM model, significantly makes the proposed In the second experiment for examining the time complexity of
method robust against missed or damaged ROIs of the input the proposed method, different number of image features is con-
images. Furthermore, it gives a high accuracy detection rate when sidered to calculate the computation time of the proposed as
the size of the trained data is suitable in the dataset. The 10 shown in Fig. 11. The proposed method used a feature vector with
patches of input image show that, the size of trained data is size 128 features due to dimensionality reduction. The running
increased to achieve about 0.95 accuracy rate. Figs. 8 and 9 show time based on 128 features is 45 s. The size of input image is
750 800. We selected ROI with size 40 40 to produce 1600 fea-
2
LivDet datasets are available to researchers at http://livdet.org/registration.php. tures. For the ROI, 5 rotated patches and 5 scaled patches are
270 D.M. Uliyan et al. / Engineering Science and Technology, an International Journal 23 (2020) 264–273
Table 1
Information of LivDet datasets 2013 and 2015 used in the proposed method.
Table 3
Performance evaluation of the proposed method on LivDet, 2013 (Biometrica and
Italdata) and LivDet, 2015 (CrossMatch).
Method
DRBM + DBM model ACC HTER Processing time per Dataset
image in seconds (s)
96.00% 3.50% 66 s Biometrica
95.00% 6.44% 85 s CrossMatch
94.50% 2.80% 45 s Italdata
Fig. 10. Computation time of the proposed method when using different number of training fingerprints.
272 D.M. Uliyan et al. / Engineering Science and Technology, an International Journal 23 (2020) 264–273
Table 4
The performance of the proposed method using LivDet 2013 with regarding four types of fabricate fingerprint materials with memory space 61.5 MB.
Table 5
Classification error rates in percentage (%) over three datasets with known spoofs.
Table 6
Comparison table shows performance evaluation of existing methods compared with [3] C. Roberts, Biometric attack vectors and defences, Comput. Sec. 26 (1) (2007)
our method on LivDet 2013. 14–25.
[4] K.W. Bowyer, Face recognition technology: security versus privacy, IEEE
Reference Techniques used ACE Technol. Soc. Mag. 23 (1) (2004) 9–19.
[5] D. Menotti et al., Deep representations for iris, face, and fingerprint spoofing
Nogueira et al. [56] Convolutional neural networks 3.9%
detection, IEEE Trans. Inform. Forensics Sec. 10 (4) (2015) 864–879.
Jiang and Xin [35] Co-occurrence matrix 11.00%
[6] M. Sajjad et al., CNN-based anti-spoofing two-tier multi-factor authentication
Gottschlich et al. [55] Histograms of gradients. 6.7%
system, Pattern Recogn. Lett. (2018).
Zhang et al. [54] Wavelet features and local binary pattern 2.1% [7] L. Ghiani et al., Livdet 2013 fingerprint liveness detection competition 2013,
The proposed method DRBM + DBM 3.6% Biometrics (ICB), 2013 International Conference on, IEEE, 2013.
[8] Q. Huang et al., An evaluation of fake fingerprint databases utilizing SVM
FerrLive þ FerrFake classification, Pattern Recogn. Lett. 60 (2015) 1–7.
ACE ¼ ð16Þ [9] T. Chugh, K. Cao, A.K. Jain, Fingerprint spoof detection using minutiae-based
2 local patches, Biometrics (IJCB), 2017 IEEE International Joint Conference on,
IEEE, 2017.
where Ferrlive is the percentage of misclassified real fingerprints
[10] R.K. Dubey, J. Goh, V.L. Thing, Fingerprint liveness detection from single image
and Ferrfake is the percentage of misclassified fake fingerprints. using low-level features and shape analysis, IEEE Trans. Inf. Forensics Secur. 11
(7) (2016) 1461–1475.
[11] C. Wang et al., A DCNN based fingerprint liveness detection algorithm with
5. Conclusion voting strategy, Chinese Conference on Biometric Recognition, Springer, 2015.
[12] Y. Gao et al., Intermediate spoofing strategies and countermeasures, Tsinghua
We proposed a novel deep learning model for examining finger- Sci. Technol. 18 (6) (2013) 599–605.
[13] D. Baldisserra et al., Fake fingerprint detection by odor analysis, International
prints based on DRBM and Deep Boltzmann Machine, which deals Conference on Biometrics, Springer, 2006.
with complex texture patterns in a robust way due to its proba- [14] Z. Akhtar et al., Evaluation of serial and parallel multibiometric systems under
bilistic multilayer architecture. In the proposed method, after spoofing attacks, IEEE Fifth International Conference: Biometrics: Theory,
Applications and Systems (BTAS), IEEE, 2012.
training a DBM, such structure has employed to extract deep fea- [15] J.J. Engelsma, K. Cao, A.K. Jain, Raspireader: open source fingerprint reader,
tures of the grayscale fingerprints. KNN classifier is applied with IEEE Trans. Pattern Anal. Mach. Intell. (2018).
the feature vectors of the ROIs extracted by the DBM to examine [16] J.-G. Wang et al., Person recognition by fusing palmprint and palm vein images
based on ‘‘Laplacianpalm” representation, Pattern Recogn. 41 (5) (2008) 1514–
spoof forgeries. The experiment results demonstrate that the Deep
1527.
learning model is robust against different kinds of spoof forgeries [17] M. Espinoza, C. Champod, P. Margot, Vulnerabilities of fingerprint reader to
such as wood glue, Gelatin and Play-Doh. Deep features are fake fingerprints attacks, Forensic Sci. Int. 204 (1–3) (2011) 41–49.
extracted from the real image of grayscale and Depth visual struc- [18] M. Drahanský, W. Funk, R. Nötzel, Liveness detection based on fine movements
of the fingertip surface, 2006, IEEE Information Assurance Workshop, 2006.
ture such as scaled and rotated patch ROIs. The performance eval- [19] C.J. Lennard et al., Fingerprints and Other Ridge Skin Impressions, CRC Press,
uation of the DRBM + DBM method achieved state-of-the-art 2016.
results in three public fingerprint recognition benchmarks. How- [20] L. Ghiani et al., Review of the fingerprint liveness detection (LivDet)
competition series: 2009 to 2015, Image Vis. Comput. 58 (2017) 110–128.
ever, our method still struggling to recognize fake fingerprints with [21] D. Yambay, Review of fingerprint presentation attack detection competitions,
unknown materials. For future work, we need to extend the Springer International Publishing, Cham, 2019, pp. 109–131.
method behavior to deal with spatio-temporal features of finger- [22] Y. Park, U. Jang, E.C. Lee, Statistical anti-spoofing method for fingerprint
recognition, Soft. Comput. 22 (13) (2018) 4175–4184.
print images to explore the liveness properties. Furthermore, we [23] J.B. Kho et al., An incremental learning method for spoof fingerprint detection,
need to explore the capability of the proposed method in reducing Expert Syst. Appl. 116 (2019) 52–64.
the time complexity of deep learning machine. [24] C. Sousedik, C. Busch, Presentation attack detection methods for fingerprint
recognition systems: a survey, IET Biom. 3 (4) (2014) 219–233.
[25] V. Mura et al., LivDet 2017 fingerprint liveness detection competition 2017,
Acknowledgments International Conference on Biometrics (ICB), IEEE, 2018.
[26] IEC, I.S., ‘‘ISO/IEC 30107-3:2016, InformationTechnology-Biometric
Presentation Attack Detection-Part 1: Framework”, in ISO/IEC 30107-1:2016
The authors are grateful to the Middle East University, Amman, 2016.
Jordan for the financial support granted to cover the publication [27] H.-U. Jang et al., Fingerprint spoof detection using contrast enhancement and
process of this research article. convolutional neural networks, International Conference on Information
Science and Applications, Springer, 2017.
[28] I. Goicoechea-Telleria et al., Presentation attack detection evaluation on
References mobile devices: simplest approach for capturing and lifting a latent
fingerprint, International Carnahan Conference on Security Technology
[1] E. Marasco, A. Ross, A survey on antispoofing schemes for fingerprint (ICCST), IEEE, 2018.
recognition systems, ACM Comput. Surv. (CSUR) 47 (2) (2015) 28. [29] K. Cao, A.K. Jain, Hacking mobile phones using 2D printed fingerprints.
[2] G. Fumera, Multimodal anti-spoofing in biometric recognition systems, in: S. Technical Report. MSU-CSE-16-2, 2016.
Marcel, M.S. Nixon, S.Z. Li (Eds.), Handbook of Biometric Anti-Spoofing: [30] A. Rattani, W.J. Scheirer, A. Ross, Open set fingerprint spoof detection across
Trusted Biometrics under Spoofing Attacks, Springer, London, 2014, pp. 165– novel fabrication materials, IEEE Trans. Inf. Forensics Secur. 10 (11) (2015)
184. 2447–2460.
D.M. Uliyan et al. / Engineering Science and Technology, an International Journal 23 (2020) 264–273 273
[31] Z. Xu, S. Li, W. Deng, Learning temporal features using LSTM-CNN architecture [43] A. Krizhevsky, I. Sutskever, G.E. Hinton, Imagenet classification with deep
for face anti-spoofing, Proceedings of the 3rd IAPR Asian Conference on Pattern convolutional neural networks, in Advances in Neural Information Processing
Recognition (ACPR), IEEE, 2015. Systems. 2012.
[32] Rattani, Ross, Automatic adaptation of fingerprint liveness detector to new [44] T. Chugh, K. Cao, A.K. Jain, Fingerprint spoof buster: use of minutiae-centered
spoof materials, IEEE International Joint Conference on Biometrics, IEEE, 2014. patches, IEEE Trans. Inf. Forensics Secur. 13 (9) (2018) 2190–2202.
[33] Xia, Z., et al., A novel weber local binary descriptor for fingerprint liveness [45] D.H. Ackley, G.E. Hinton, T.J. Sejnowski, A learning algorithm for Boltzmann
detection, In IEEE Transactions on Systems, Man, and Cybernetics: Systems, machines, Cognit. Sci. 9 (1) (1985) 147–169.
2018. [46] G.E. Hinton, R.R. Salakhutdinov, Reducing the dimensionality of data with
[34] P.D. Lapsley, et al., Anti-fraud biometric scanner that accurately detects blood neural networks, Science 313 (5786) (2006) 504–507.
flow. 1998, Google Patents. [47] P. Smolensky, Information processing in dynamical systems: Foundations of
[35] Y. Jiang, X. Liu, Spoof fingerprint detection based on co-occurrence harmony theory, 1986, Univ at boulder dept of computer science colorado.
matrix, Int. J. Signal Process. Image Process. Pattern Recogn. 8 (8) (2015) [48] V. Nair, G.E. Hinton. Implicit mixtures of restricted Boltzmann machines, in
373–384. Advances in Neural Information Processing Systems, 2009.
[36] L. Ghiani, G.L. Marcialis, F. Roli, Fingerprint liveness detection by local phase [49] G.E. Hinton, S. Osindero, Y.-W. Teh, A fast learning algorithm for deep belief
quantization, Proceedings of the 21st International Conference on Pattern nets, Neural Comput. 18 (7) (2006) 1527–1554.
Recognition (ICPR2012), IEEE, 2012. [50] M. Welling, G.E. Hinton, A new learning algorithm for mean field Boltzmann
[37] A. Abhyankar, S. Schuckers, Fingerprint liveness detection using local ridge machines, International Conference on Artificial Neural Networks, Springer,
frequencies and multiresolution texture analysis techniques, International 2002.
Conference on Image Processing, IEEE, 2006. [51] H. Yu, J. Yang, A direct LDA algorithm for high-dimensional data-with
[38] S.B. Nikam, S. Agarwal, Fingerprint liveness detection using curvelet energy application to face recognition, Pattern Recogn. 34 (10) (2001) 2067–2070.
and co-occurrence signatures, Proceedings of the fifth international conference [52] M. Muja, D.G. Lowe, Scalable nearest neighbor algorithms for high dimensional
on computer graphics, imaging and visualisation, IEEE, 2008. data, IEEE Trans. Pattern Anal. Mach. Intell. 11 (2014) 2227–2240.
[39] P. Coli, G.L. Marcialis, F. Roli. Power spectrum-based fingerprint vitality [53] L. Ghiani et al., Livdet 2013 fingerprint liveness detection competition 2013,
detection, in 2007 IEEE Workshop on Automatic Identification Advanced International Conference on Biometrics (ICB), IEEE, 2013.
Technologies, 2007. IEEE. [54] Y. Zhang et al., Fake fingerprint detection based on wavelet analysis and local
[40] T. de Freitas Pereira et al., LBP TOP based countermeasure against face binary pattern, Chinese Conference on Biometric Recognition, Springer, 2014.
spoofing attacks, Asian Conference on Computer Vision, Springer, 2012. [55] C. Gottschlich et al., Fingerprint liveness detection based on histograms of
[41] Z. Akhtar et al., Evaluation of multimodal biometric score fusion rules under invariant gradients, Biometrics (IJCB), 2014 IEEE International Joint
spoof attacks, Proceedings of the 5th IAPR International Conference on Conference, Citeseer, 2014.
Biometrics (ICB), IEEE, 2012. [56] R.F. Nogueira, R. de Alencar Lotufo, R.C. Machado, Evaluating software-based
[42] J. Galbally et al., A high performance fingerprint liveness detection method fingerprint liveness detection using convolutional networks and local binary
based on quality related features, Future Gener. Comput. Syst. 28 (1) (2012) patterns, Biometric Measurements and Systems for Security and Medical
311–321. Applications (BIOMS) Proceedings, 2014 IEEE Workshop, IEEE, 2014.