B IOMETRIC recognition is one of the means that can be the obtained - - PDF document

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2755771, IEEE Access

Access-2017-05959 1

 Abstract—Handwritten Signature Recognition is a biometric mode that has started to be deployed. Therefore, it is necessary to analyse the robustness of the recognition process against Presentation Attacks, to find its vulnerabilities. Using the results

• f a previous work, the vulnerabilities are detected and two

Presentation Attack Detection techniques have been implemented. With such implementations, a new evaluation has been performed, showing an improvement in the performance. Error rates have been lowered from about 20% to below 3% under operational conditions. Index Terms—Biometrics, Dynamic Analysis, Handwritten Signature, Presentation Attack Detection, Robustness Evaluation

• I. INTRODUCTION

IOMETRIC recognition is one of the means that can be used to identify or authenticate a citizen in an automatic

• way. Within the different biometric traits that can be used, one

that has a direct application in many scenarios is handwritten

• signature. But regarding biometric recognition, handwritten

signature involves two different biometric modes: the use of the static information of the signature (i.e. the graph drawn), and the use of the dynamic information of the act of signing (e.g. timing evolution of the position of the writing stylus or the pressure applied at each moment of the signature). Different studies have shown that, using dynamic signature, a better performance is obtained than using static signature, even one

• rder of magnitude better [1]. Therefore, this paper is focused
• n Dynamic handwritten Signature Verification (DSV).

Several authors have worked in DSV improving its performance using different algorithms [2]. One of the most used algorithms is the use of Dynamic Time Warping (DTW) [3], which has also achieved the best results in some public competitions [4][5]. Nevertheless, there is room for further research in improving the application of DTW to dynamic handwritten signature, such as using the results obtained in

• ther generic DTW works [6].

This biometric mode has been proven to be applicable in real life, even using different kinds of devices or writing elements (e.g. stylus or finger) [7], different stylus technologies [8], or even under stress conditions [9]. Novel implementations such as signing in-air have also been developed by other authors

Submitted 02/08/2017.

• R. Sanchez-Reillo, H. C. Quiros-Sandoval, I. Goicoechea-Telleria, and W.

Ponce-Hernandez are with the Carlos III University of Madrid, Avda. de la

[10]. But when an authentication technique is ready for being deployed, it is essential to evaluate its vulnerabilities and solve

• them. This has been done with other modalities, such as

fingerprint [11] or face [12]. In the case of DSV, the major vulnerability is the one related to Presentation Attacks (PA), in particular, forgeries. This paper is related to the creation of Presentation Attack Detection (PAD) mechanisms and their evaluation, so as to determine the level of robustness achieved. Therefore, this paper will first explain and summarize the previous works from the authors related to the evaluation of the robustness. This is detailed in section II. After that, section III will analyse those previous results in order to determine where the major vulnerabilities can be found, and detail a strategy to cover them. Section IV will provide a couple of PAD mechanisms, showing the obtained results in Section V. The paper will finish with the conclusions and future working lines proposed.

• II. BACKGROUND

In order to understand this work and its impact, it is necessary to revisit a previous work from the authors. In such work [13], authors developed an evaluation platform in order to test the robustness of handwritten signature biometrics against

• forgeries. This clause summarizes the evaluation methodology,

as well as the results obtained.

• A. Evaluation Methodology

The evaluation platform was developed following all current international standards, such as the data format in ISO/IEC 19794-7 [14], the particular evaluation conditions for handwritten signature described in ISO/IEC 19795-3 [15], and the recent standard on the evaluation of PAD in ISO/IEC 30107-3 [16]. Such evaluation platform exploited the level of knowledge gained by the forger as he/she learns about the signature to be

• forged. So, the forger performs the training on the target

signature following 11 levels of knowledge, as it is represented in Fig.1. The first 7 levels represent Laboratory Conditions, where a set of tools are available to the forger. In the first level, the forger does not know anything about the signature to be forged

Universidad, 30, 28911, Leganes, Spain. (e-mail: {rsreillo, hquiros, igoicoec, and wponce}@ing.uc3m.es).

Improving Presentation Attack Detection in Dynamic Handwritten Signature Biometrics

Raul Sanchez-Reillo, Member, IEEE, Helga C. Quiros-Sandoval, Ines Goicoechea-Telleria, and Wendy Ponce-Hernandez

B

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2755771, IEEE Access

Access-2017-05959 2 (i.e. this represents a zero-effort attack). In Level 2, the forger sees the image of the signature for only 5 seconds. Level 3 allows the forger to see the graph of the signature at all times, even facilitating a way of carbon copying the signature in Level

• 4. Levels 2 to 4 exploit the knowledge about the static

information of the signature. Level 5 starts with the dynamic information, showing the forger a single reproduction of the signature while it was

• written. Level 6 provides the forger with a signature player,

with which he/she can reproduce the execution of the signature slower, faster, forward, backwards, etc. The last laboratory level merges the signature player with the carbon copy facility.

• Fig. 1. Knowledge-based attack levels used by the Presentation

Attack Evaluation Platform developed in [13]

Once finished with the Laboratory Conditions, all tools are removed from the forger, and he/she is asked to forge the signature by heart. This is done immediately after finishing Level 7. Then, the user is asked to wait for one hour before trying it again, becoming Level 9. Furthermore, the forger is asked to forge it again after 4 more hours, so as to force him/her to perform any other kind of activity, such as having lunch or

• dinner. This is Level 10. Finally, the forger is sent back home,

and asked to try the forgery again after sleeping (i.e. after another 12 hours). These last 4 levels are considered to be the simulation of Operational Conditions. The signatures acquired with the platform will be used as PA, and they were taken using a desktop computer application and a Wacom STU-500 signing pad. In addition, operational levels were also acquired using mobile phones and tablets, both using a native stylus and the finger to sign. The original signatures from the users are considered bona- fide signatures and used as a baseline for enrolment and baseline algorithm performance. Bona-fide signatures were taken with all the devices used for forgeries (i.e. Wacom STU, mobile phones and tablets).

• B. Previous Results

Using such platform, a Dynamic Time Warping (DTW) based dynamic handwritten signature verification solution was

• evaluated. This solution is detailed in [7], which, at the same

time, is based on [17]. Within the published evaluation, an

• perational scenario was approached. For example, the

enrolment took only the first 5 bona-fide signatures accepted by the user. For each attack level, a set of 10 forgeries were acquired.

• Fig. 2. Normalized distributions of scores. Green lines represent

mated comparisons, red lines represent non-mated, and black lines represent attacks. Figures are, from left to right and from top to bottom, levels 1 to 11. Vertical axis represents normalized distribution in a scale from 0 to 100. Horizontal axis represents the comparison score, in terms of distance.

• Fig. 2 illustrates the behaviour of the baseline algorithm

when attacks are added. The illustration is done by using the normalized distribution curves, both for mated (i.e. intra-class distribution) in green, and non-mated (i.e. inter-class distribution) in red. In addition, the black line represents the score distribution of the attacks, and a figure is shown for each

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2755771, IEEE Access

Access-2017-05959 3

• f the levels.

As it can be easily seen, as the knowledge level increases, the

• verlap between the black line and the green line becomes
• significant. In terms of error rates, Fig. 3 shows the main results.
• Fig. 3. DET curves showing the performance of the system at levels

4, 7, 8 – 11, compared to the baseline behaviour with no attacks

Analysing the results, the Equal Error Rate (EER) between the Imposter Attack Presentation Match Rate (IAPMR) and the False Non-Match Rate (FNMR) for the bona-fide are: 1.72% (baseline), 23.35% (Level 4), 33.43% (Level 7), 19.27% (Level 8), 18.27% (Level 9), 16.93% (Level 10) and 18.21% (Level 11). Similar results for the operational conditions are found where mobile devices are used, either with a stylus or with the finger [13]. Therefore, the impact of well-trained forgeries is extremely significant for this particular algorithm. This is the reason why, in this paper, this fact is analysed and several solutions are proposed.

• III. VULNERABILITY ANALYSIS OF PREVIOUS RESULTS

There are several conclusions that can be obtained by analysing the results obtained in [13] and summarized in the previous section. The first one is the ability of the forger to remember the signature even one day later. As it can be seen, error rates in levels 8 to 11 are lower than those on levels 4 and 7, but still much higher than those of the baseline. In all cases the forger success rate is close to a 20%, or much higher depending on the threshold used by the system implementer. A second very important outcome is the fact that, by only knowing the static information of the signature, the error levels increase very significantly, going up to 23%. This contradicts the idea that a dynamic signature algorithm becomes more robust as the dynamic information is considered within the

• decision. In fact, the increase in the error rates when knowing

the dynamic information is not as high as expected. This questions if the dynamic recognition algorithm really uses the dynamic information. The reason behind this behaviour is that the dynamic warping

• f the DTW algorithm works very well for fitting the different

signatures from the same signer (i.e. narrowing the intra-class variability). But within that process, several dynamic-related information is removed. In particular, one of the effects that the implementation of the DTW causes is the re-scaling of the time signals to a common

• duration. This allows to wrongly accept a forgery that

reproduces the same variation of the signals, but performed very slowly. Also, depending on the DTW implementation, another information lost is the number of strokes used for signing, and the length and significance of those strokes. And due to the fact that a user may vary his number of strokes among signatures, assigning strokes as to perform a stroke-based DTW is not a trivial task.

• IV. PROPOSED PAD MECHANISMS

With all these considerations, a potential way to reduce the success of forgeries is to add dynamic information to the comparison process. This section proposes 2 additional metrics to be added to the comparison process as to reject forgeries.

• A. Number of Strokes

The first metric is the number of strokes. As already mentioned, the number of strokes that a user draws during signing may vary, but it is most of the times within an interval. Determining the number of strokes in an ISO/IEC 19794-7 biometric record is as simple as analysing the S channel. The starting point of a stroke is determined by the transition from 0 to 1. The number of transitions represents the number of strokes. The variation of the number of strokes depends also on the number of strokes executed. In other words, a signature with a low number of strokes may present a low number of different strokes finally drawn. But if the signature typically has a large number of strokes, the diversity of number of strokes will be

• larger. Therefore, the threshold to detect a signature as a

potential forgery is set as a percentage of the average number

• f strokes during enrolment.
• B. Signing Time

The second metric will be the time that the user needs to sign. Again, this time varies from signature to signature, and is also dependent on the length of complexity of the signature. But the variation is expected to be limited for each of the users. This metric has an advantage from the usability point of view: the user may understand a rejection if the signature has taken longer

• r shorter than usual.

But the variations when the signing device is in the air are much larger than the ones when the signing device is on the

• surface. Therefore, the time used as a basis for this metric is the

time while signing, removing the time when the pointing device is in the air. The resulting time is the sum of the time needed for each of the strokes. The basis for the comparison will be the average of that time during enrolment.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2755771, IEEE Access

Access-2017-05959 4

• V. PAD EVALUATION RESULTS

These two additional metrics were added as initial checks before applying the baseline algorithm. For illustration purposes, the system will behave in a way that if any of the checks determine that the signature may be a forgery, a maximum distance will be assigned in the comparison. This will impact even the baseline behaviour, as many of the non- mated comparison may be considered as forgeries. Fig. 4 shows such a behaviour. It is important to note that the results shown here are based on the same dataset and enrolment conditions as the ones in [13].

• Fig. 4. Impact of a PAD mechanism in the normalized curves for

distance distribution (green – mated; red – non-mated). Axes represent the same values as in Fig. 2.

Initially it may seem that the performance of the algorithm is much better, as both distribution curves seem to be much more

• differentiated. Unfortunately, some of the mated comparisons

are also considered as forgeries, and therefore there is still an

• verlap between both distributions.

In other words, the addition of a PAD mechanism always impacts the baseline performance, and sometimes could even cause a worse performance. For example, in our case, as it will be seen in Table I, the baseline EER will raise from 1.72% to

• ver 2.1%, and such increase is due to the mated comparisons

considered as potential forgeries.

• A. Results based on Number of Strokes

When applying the first of the metrics, the behaviour of the system can be seen in the normalized distribution graphs of Fig.

• 5. In each of the 11 plots in Fig. 5, there are 5 curves

represented: the mated baseline (i.e. the one without PAD) in dotted green line, the new mated curve (i.e. the one with PAD) in continuous green line, the non-mated baseline in dotted red line, the new non-mated curve in continuous red line, and the forgeries in black. Both mated curves are alike (except for the small peak in the maximum distance for the new non-mated curve), therefore, only one green line is distinguished. The new non-mated curve is mostly located at the maximum distance (as in Fig. 4), so only the non-mated baseline is visible at other

• scores. Forgeries distributions are the ones that vary from plot

to plot, representing the behaviour for all 11 levels. It can be seen how the attacks are well separated from mated distribution in Level 1, but as the level increases, the overlap of both curves starts to become significant. It is true that up to Level 4, the overlap is lower than in the baseline system. When the forger starts to learn the dynamics of the signature, the PAD mechanism starts to lose its discriminant power. But when analysing the EERs obtained (see Table I), the system achieves worse results in all cases, as this metric has a major impact in the intra-class distribution. Graphically, major results can be seen in Fig. 6. Therefore, this metric by itself does not provide an improvement in the performance.

• Fig. 5. Normalized distributions of scores with PAD based on

number of strokes. Green lines represent mated comparisons, red lines represent non-mated, and black represent attacks. Figures are, from left to right and from top to bottom, levels 1 till 11. Axes represent the same values as in Fig. 2.

• B. Results based on the Addition of Signing Time

Adding the signing time to the previous PAD mechanism, the results that can be seen in Fig. 7 (distribution curves) and Fig. 8 (DET curves) are achieved. As it can be extracted from the visual inspection of Fig. 7, this new PAD mechanism seems to work much better, as in all 11 levels, the attack distribution curve (i.e. the black one) does not show much of an overlap with the mated curves.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2755771, IEEE Access

Access-2017-05959 5

• Fig. 6. DET curves showing the performance of the system with PAD

based on strokes, at levels 4, 7, 8 – 11, compared to the baseline and the bona-fide behaviour

• Fig. 8 shows the DET curves for all levels, including also the

baseline (in green) and the behaviour of the system with PAD with bona-fide signatures. All intersections with the EER diagonal are below 5%. The third column in Table I shows the EER values obtained in this case for each of the levels. The highest EER is 3.1%, much lower than the behaviour of the system without PAD, and quite close to the performance

• btained using only bona-fide signatures. Therefore, the

detection of the forgeries is achieved by a higher discrimination

• f the inter-class variability.
• C. Comparison with State-of-the-Art

A direct comparison of these results with the State of the Art is nearly impossible, as there is not one work published using this methodology to test the robustness against forgeries. There is also no public database with the same characteristics than the

• ne created for this study. But an indirect comparison can be

done by analysing the results with one of the most used databases in the literature: the MCYT database [18]. Using MCYT for the evaluation, the baseline algorithm used for this study presents an EER with bona-fide signatures (also known as random-forgeries) of 0.54%. With skilled-forgeries, the performance of the baseline algorithm reached 3.6%. These results are in the same order of magnitude of those reported in [1] and [17].

• Fig. 7. Normalized distributions of scores with PAD adding timing.

Green lines represent mated comparisons, red lines represent non- mated, and black represent attacks. Figures are, from left to right and from top to bottom, levels 1 to 11. Axes represent the same values as in Fig. 2.

As it can be seen, the performance results between the ones

• btained with the skilled forgeries in MCYT and the

progressively training samples in this study differ very

• significantly. This provides a new evidence of the importance

and validity of the developed evaluation platform. At the same time, it also shows that the PAD techniques reported in this paper provide error rates in the same order of magnitude of the

TABLE I EERS (IN %) ACHIEVED IN EACH IMPLEMENTATION Level Baseline Strokes Strokes + Timing Bona-fide 1.7 2.1 2.1 L1 1.6 2.2 2.1 L2 6.8 8.4 3.2 L3 12.9 13.6 1.8 L4 23.4 25.0 1.8 L5 17.7 19.2 2.1 L6 19.5 20.7 2.0 L7 33.4 35.7 1.7 L8 19.3 21.4 2.3 L9 18.3 18.6 3.1 L10 16.9 17.7 3.1 L11 18.2 18.0 2.6

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2755771, IEEE Access

Access-2017-05959 6

• nes published in the State-of-the-Art, with a much more

demanding testing database.

• Fig. 8. DET curves showing the performance of the system with PAD

adding timing, at levels 4, 7, 8 – 11, compared to the baseline and the bona-fide behaviour

• VI. CONCLUSION

The present paper has analysed the vulnerabilities shown in a previous work related to the evaluation of a DTW-based handwritten signature recognition solution. The vulnerabilities detected have provided ideas to design Presentation Attack Detection techniques that could improve the performance against forgeries. Two of those metrics have been reported, being the first one not effective, but becoming effective when combined with the second one. The improvement achieved is close to one order of magnitude, reducing the error rates from a percentage close to 20%, to error rates below 3%. Further work should explore other PAD mechanisms, as well as study the user-system interaction when rejections related to a potential forgery are implemented. REFERENCES

[1]

• R. Plamondon and S. N. Srihari, “On-line and off-line handwriting

recognition: A comprehensive survey,” IEEE Trans. Pattern Anal.

• Mach. Intell., vol. 22, no. 1, 2000.

[2]

• D. Impedovo and G. Pirlo, “Automatic signature verification: The

state of the art,” IEEE Trans. Syst. Man Cybern. Part C Appl. Rev.,

• vol. 38, no. 5, 2008.

[3]

• M. Faundez-Zanuy, “On-line signature recognition based on VQ-

DTW,” Pattern Recognit., vol. 40, no. 3, pp. 981–992, Mar. 2007. [4] D.-Y. Yeung et al., “SVC2004: First International Signature Verification Competition,” in Biometric Authentication: First International Conference, ICBA 2004, Hong Kong, China, July 15- 17, 2004. Proceedings, D. Zhang and A. K. Jain, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2004, pp. 16–22. [5]

• N. Houmani et al., “BioSecure signature evaluation campaign

(BSEC’2009): Evaluating online signature algorithms depending on the quality of signatures,” Pattern Recognit., vol. 45, no. 3, pp. 993– 1003, 2012. [6]

• E. Keogh and C. A. Ratanamahatana, “Exact indexing of dynamic

time warping,” Knowl. Inf. Syst., vol. 7, no. 3, pp. 358–386, Mar. 2005. [7]

• R. Blanco-Gonzalo, R. Sanchez-Reillo, O. Miguel-Hurtado, and J.

Liu-Jimenez, “Performance evaluation of handwritten signature recognition in mobile environments,” IET Biometrics, vol. 3, no. May 2013, pp. 139–146, 2013. [8]

• R. Blanco-Gonzalo, L. Diaz-Fernandez, O. Miguel-Hurtado, and R.

Sanchez-Reillo, “Usability Evaluation of Biometrics in Mobile Environments,” in The 6th International Conference on Human System Interaction (HSI), 2013, vol. 300, pp. 123–128. [9]

• R. Blanco-Gonzalo, R. Sanchez-Reillo, and N. Poh, On the effect of

time efficiency in DSV under stress, vol. 2014–Octob, no. October. 2014, pp. 1–5. [10]

• G. Bailador, C. Sanchez-Avila, J. Guerra-Casanova, and A. De

Santos Sierra, “Analysis of pattern recognition techniques for in-air signature biometrics,” Pattern Recognit., vol. 44, no. 10–11, 2011. [11]

• I. Bhardwaj, N. D. Londhe, and S. K. Kopparapu, “Study of

Imposter Attacks on Novel Fingerprint Dynamics Based Verification System,” IEEE Access, vol. 5. pp. 595–606, 2017. [12]

• J. Galbally, S. Marcel, and J. Fierrez, “Biometric Antispoofing

Methods: A Survey in Face Recognition,” IEEE Access, vol. 2. pp. 1530–1552, 2014. [13]

• R. Sanchez-Reillo, H. C. Quiros-Sandoval, J. Liu-Jimenez, and I.

Goicoechea-Telleria, “Evaluation of strengths and weaknesses of dynamic handwritten signature recognition against forgeries,” in Proceedings - International Carnahan Conference on Security Technology, 2016, vol. 2015–Janua. [14] “ISO/IEC 19794-7:2014 Information technology - Biometric data interchange formats - Part 7: Signature/sign time series data.” ISO/IEC JTC1, 2014. [15] “ISO/IEC TR 19795-3:2007 Information technology - Biometric performance testing and reporting - Part 3: Modality-specific testing.” ISO/IEC JTC1, 2007. [16] “ISO/IEC 30107-3 Information technology - Biometric presentation attack detection - Part 3: Testing and reporting.” ISO/IEC JTC1, 2017. [17]

• J. M. Pascual-Gaspar, V. Cardenoso-Payo, and C. E. Vivaracho-

Pascual, Practical on-line signature verification, vol. 5558 LNCS. 2009. [18]

• J. Ortega-Garcia et al., “MCYT baseline corpus: A bimodal

biometric database,” IEE Proc. Vision, Image Signal Process., vol. 150, no. 6, 2003.

Raul Sanchez-Reillo (M’97) Prof. Raul Sanchez-Reillo obtained his PhD in Telecommunication Engineering in 2000, by Universidad Politecnica de Madrid. He is now Associate Professor at Carlos III University of Madrid (UC3M). He is also the Head of the University Group for Identification Technologies (GUTI). His R&D group is involved in project development related to identification technologies, either by the user of secure elements (such as smartcards) and/or by using

• Biometrics. In addition to R&D activities, he has also managed

projects concerning a broad range of applications, from Social Security Services till Financial Payment Methods. He has taken part in European Projects like eEpoch, BioSec, BEST Network, EKSISTENZ, MobilePass and ORIGINS, being WP leader in most of them. He is member of SC17, SC27 and SC37 Standardization Committees. He holds the Spanish Chair in SC17 and the Spanish Secretariat in SC37, as well as the Secretariat of the international working groups SC17 WG11 and SC37 WG2. In 2009 he founded IDTestingLab, an Evaluation Laboratory for Identification Products. He is member of several ad-hoc groups such as the Biometric Vulnerability Assessment Expert Group (BVAEG).

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2755771, IEEE Access

Access-2017-05959 7 Helga C. Quiros-Sandoval graduated as Electrical Engineer with mention in Electronics and Communications by University of Carabobo in Venezuela, actually studying a PhD at de University Carlos III of Madrid, based on Robustness

• f Dynamic Signature Biometrics against

Fraudulent Attacks. From 2014 she has been researching at the University Group for Identification Technologies, developing her first project based on the viability study of the fraudulent attacks over handwritten signature systems, with which she obtained her Master degree in Electronic Systems Engineering and Applications. Ines Goicoechea-Telleria obtained her Bachelor's Degree in Industrial Electronics and Automation at University Carlos III of Madrid (UC3M) in 2014 and a Master's Degree in Electronic Systems and Applications Engineering in 2015. She is currently working at the Electronics Technology Department in University Carlos III of Madrid (UC3M), as part of the University Group for Identification Technologies (GUTI). She joined GUTI in October 2014 and has been working on Presentation Attack Detection since then. She is now currently doing her PhD on the Evaluation of Presentation Attack Detection in the context of Common Criteria and is a member of ISO/IEC JTC1 SC27 and SC37, and CEN/TC 224 WG18. Wendy Ponce-Hernandez graduated as Telecommunication Engineer by Technological University of Havana Jose Antonio Echeverria, Cuba. In 2015, she

• btained a Master Degree in Electronic

Systems Engineering at Carlos III University of Madrid. From 2015 she has been researching in Biometric Template Protection at the University Group for Identification Technologies at UC3M, where currently she studies her PhD.