|
Databases
Please note the license must be signed by a permanent faculty or research member from your university, and a scanned copy must be attached to the application.
Raw biometric data
-
MobileB2C_BehavePassDB: Mobile HCI data (keystroke, touch gestures, background sensors) for behavioral biometrics used in the MobileB2C ongoing competition.
-
DeepSignDB: Database comprises a total of 1526 users from four different popular databases.
-
MobileTouchDB: On-line character database performed by 217 users, using 94 different smartphone models in an unsupervised scenario.
-
KBOC_DB: Keystroke dataset of the KBOC competition with more than 300 users and 24 keystroke samples acquired during 4 sessions.
-
TouchDB: public benchmark to evaluate swipe biometrics extracted from human-device interaction with touch screens.
-
e-BioDigit Database: Handwritten numerical digits database acquired using a Samsung Galaxy Note 10.1 general purpose tablet for a total of 93 users. All samples were acquired using the finger touch as input.
-
BIOGIGA Database: The baseline corpus of BioGiga consists of synthetic images at 94 GHz (within millimetre waves) of the body of 50 individuals. The images are the result of simulations carried out on 3D-corporal models at two types of scenarios (outdoors, indoors) and with two kinds of imaging systems (passive and active). These corporal models were previously generated based on body measures taken from real subjects.
-
e-BioSign-DS1-Signature DB: Signature database acquired from 5 different COTS devices in total, considering both pen stylus and also the finger. Signatures were collected in two sessions for 65 subjects.
-
BiosecurID-SONOF DB: Two datasets containing real and synthetic on-line and off-line signatures for 132 users, with 16 genuine signatures and 12 skilled forgeries per user.
-
ATVS Keystroke DB: A dataset containing keystroking data of 63 users, including 12 genuine and 12 skilled impostor samples per user.
-
ATVS-SG_NOTE: Signature database captured using a Samsung Galaxy Note device with 25 users and 20 signatures per user.
-
xLongSignDB (Extended version of the ATVS On-Line Signature Long-Term database): The dataset comprises the on-line signature data of the 29 common users to the BiosecurID and the Biosecure databases.
-
ATVS-FakeIris Database (ATVS-FIr DB): A dataset containing 1,600 real and fake fingerprint images specifically thought to assess the vulnerability of
iris-based recognition systems to direct attacks and to evaluate the performance of liveness detection methods. Fake samples were captured from high quality printed iris images.
-
Swansea Footstep Biometric Database (SFootBD): Footstep database containing pressure information over time for two 88-sensor arrays. It contains 9,990 stride footstep (right and left) signals from 127 persons specifically thought to assess the performance of footsteps as a biometric.
-
ATVS-SyntheticSignature Database (ATVS-SSig DB): Two datasets containing 17,500 highly-realistic fully-synthetic on-line signatures specifically thought to assess the performance of signature-based recognition systems, or for security evaluation purposes.
-
ATVS-FakeFingerprint Database (ATVS-FFp DB): Two datasets containing over 4,500 real and fake fingerprint images specifically thought to assess the vulnerability of fingerprint-based recognition systems to direct attacks and to evaluate the performance of liveness detection methods. Fake samples were captured from gummy fingers generated both with and without the cooperation of the user. Three different sensors were used to acquire the database: flat-optical, flat-capacitive, and sweeping-thermal.
-
ATVS-DooDB: Finger-drawn graphical password database: Doodles and pseudo-signatures; 100 users
-
FVC2006 Fingerprint Database (FVC2006): the database used in the Fingerprint Verification Competition 2006
-
MCYT Bimodal Biometric Database (MCYT-Signature-100): on-line signature; 100 users
-
MCYT Bimodal Biometric Database (MCYT-SignatureOff-75): off-line signature; 75 users
-
MCYT Bimodal Biometric Database (MCYT-Fingerprint-100): fingerprint; 100 users
-
GANDiffFace: source code for the synthesis of face images with realistic variations.
Processed biometric data
-
TouchDB_Benchmark: This code includes a public benchmark to evaluate swipe biometrics extracted from human-device interaction with touch screens.
-
LFW Soft Biometrics Database: This database contains a collection of soft biometrics extracted from the LFW database. We include two versions of the database: 1) manual annotations, and 2) automatic annotations using two COTS: Face++ and Microsoft API. The database also provides face recognition scores from the 10-folds from view 2 of the LFW database using features extracted from the VGG-16 pre-trained model.
-
BiosecurID-SGlobalLocalFeat DB: The BiosecurID-SGlobalLocalFeat DB contains two complementary feature datasets, extracted from the BiosecurID signature corpus: Global features containing a fixed-length representation of the on-line signatures, and Local functions containing a variable-length representation of the on-line signatures, based on time sequences.
-
Guardia Civil Database (GCDB_Features): A dataset from real forensic casework containing 268 latent fingerprint minutia templates and their corresponding 268 mated tenprint fingerprint minutia templates, including rare minutiae.
-
SCfaceDB Facial Landmarks Database (SCfaceDB Landmarks): A dataset containing 21 facial landmarks (from 4,160 face images) from 130 users manually annotated by a human operator. The dataset is especially suited to perform experiments related to facial region extraction and face recognition.
-
Tunnel Database Soft Biometric (TunnelDBSoftBio): A dataset containing soft biometric signals (from 23 physical trait labels) of 58 users manually annotated by 10 different annotators. The dataset is especially suited to perform experiments related to soft biometrics and face recognition.
-
MCYT-SCORES Database (MCYT-SCORES): Three datasets of bimodal matching scores (signatures and fingerprints from 75 subjects of the MCYT database), together with scalar fingerprint quality measures labelled by a human expert, both in MATLAB and ASCII format.
BiDA Lab -
Biometrics and Data Pattern Analytics Research Group » Databases » BiosecurID_SONOF_DB
|
INSTRUCTIONS FOR DOWNLOADING
BiosecurID-SONOF DB
-
Download license agreement,
send by email one signed and scanned copy to atvs uam.es.
-
Send an email to atvsuam.es, as follows:
Subject:
[DATABASE download: BiosecurID-SONOF DB]
Body: Your name, e-mail, telephone
number, organization, postal mail, purpose for which you will use
the database, time and date at which you sent the email with the
signed license agreement.
-
Once the email copy of the license
agreement has been received at ATVS, you will get an email with a
username, a password, and a time slot to download the database.
-
Download the database, for which you will need to provide the
authentication information given in step 4. After you finish the
download, please notify by email to atvsuam.es that you have
successfully completed the transaction.
-
For more information, please contact: atvsuam.es
DESCRIPTION OF BiosecurID-SONOF DB
The BiosecurID-SONOF DB contains two complementary datasets:
Real dataset: containing on-line and off-line versions of the exact same signatures. Synthetic dataset: containing off-line signatures generated according to the method described in [PR2015] based on the dynamic signatures of the real database.
Some examples of the data that can be found in the database are shown in Fig. 3.
The real dataset is a subcorpus of the signature data contained in the BiosecurID multimodal database. BiosecurID was acquired in five different Spanish universities and comprises eight different biometric traits captured in four sessions over a six month time span [PAA2010]. The subcorpus included in the BiosecurID-SONOF Real dataset is the signature data corresponding to the 132 users acquired at the Universidad Autonoma de Madrid.
The BiosecurID-Signature UAM subcorpus comprises 132 users, with 16 genuine signatures (four per session) and 12 skilled forgeries (three per session) for every subject. Hence, the database contains the on-line and off-line data of 16x132=2,112 genuine signatures and of 12x132=1,584 skilled forgeries.
Handwritten signatures were acquired with the Intuos3 A4/Inking pen tablet placing a predefined paper template over the digitizing device as shown in Fig 1. The users were told to sign inside a delimited grid in order to reduce the rotation and size variations (25 mm x 120 mm). Signatures were performed on the marked area with a special inking pen which also captured the x and y trajectories and the pen pressure during the signing process, with a sampling frequency of 100 Hz. This way, both versions, dynamic and static, of the same samples were captured simultaneously. In order to obtain the final off-line digitized samples, the grid-templates used to capture the static signatures were scanned at 600 dpi into png grey level files, which were then processed to automatically segment the signature images, stored with the same codename as their on-line versions.
Consequently, the database contains the off-line (on paper) and on-line versions of the exact same real signatures. Genuine and skilled forgery real samples of the same user are shown in the first two rows of Fig. 3, where both the dynamic and static versions of the same signatures are depicted.
Fig. 1: Diagram of the BiosecurID DB acquisition process. Users signed on a paper template that limited the scaling and rotation variability, placed over a digitizing tablet. This way, the on-line and off-line versions of the same signature were acquired simultaneously (figure extracted from [PR2015]).
The synthetic off-line data was generated taking as input the on-line real signatures of the BiosecurID-SONOF Real Dataset. That is, for each real on-line signature in the BiosecurID-SONOF Real Dataset (genuine or skilled forgery), its off-line synthetic version is produced following the methodology described in [PR2015]. Therefore, the synthetic off-line dataset presents exactly the same structure as the real version, that is: 4 sessions, 132 users, 4 genuine signatures and 3 skilled forgeries per session and user.
As described in [PR2015] and shown in Fig. 2, for each real on-line signature, two different synthetic images are produced:
Last row in Fig. 3 shows the synthetic static samples corresponding to the three real signatures depicted in the first two rows. Synthetic signatures are defined by two different images: Ienhanced (third row, top), which incorporates pressure and speed information from the real dynamic signature; and Ipen-ups (third row, bottom), obtained from the signature trajectory during pen-ups.
Fig. 2: Diagram of the enhanced off-line signature generation approach described in [PR2015].
Fig. 3: Real on-line, real off-line and synthetic off-line versions of typical signature examples that can be found in the BiosecurID-SONOF DB. Two genuine samples (first two columns) and a skilled forgery (last column) of the same user are shown. On-line samples are depicted with their corresponding time functions (x and y trajectories and pressure function p). Synthetic samples were generated following the method described in [PR2015]. Each synthetic signature is defined by two images: Ienhanced (third row, top) and Ipen-ups (third row, bottom).
FILES FORMAT
On-line signatures are stored in MATLAB .mat files which contain three vectors [x, y, p] each of them corresponding to each of the three time functions defining each signature (i.e., horizontal and vertical coordinates and pressure signal).
Off-line signatures (both real and synthetic) are stored in regular 600dpi image .png grey level files.
NOMENCLATURE
There is a slight difference between the naming applied to on-line and off-line files.
The nomenclature followed to name the on-line signature files is as follows: uXXXX_sYYYY_sgZZZZ
XXXX: is the number of the user [1001, 1002 ... 1132] YYYY: is the number of the session [0001, 0002, 0003, 0004] ZZZZ: is the number of the sample [0001, 0002, 0003, 0004, 0005, 0006, 0007]
Signatures [1, 2, 6, 7] of each session are genuine samples.
Signatures [3, 4, 5] of each session are skilled forgeries.
The nomenclature followed to name the on-line signature files is as follows: uXXXX_sYYYY_sgZZZZA
XXXX: is the number of the user [1001, 1002 ... 1132] YYYY: is the number of the session [0001, 0002, 0003, 0004] ZZZZ: is the number of the sample [0001, 0002, 0003, 0004, 0005, 0006, 0007] A: can take the values "g" for genuine signatures (samples [1, 2, 6, 7]) and "f" for skilled forgeries (samples [3, 4, 5]).
REFERENCES
For further information on the database we refer the reader to (all these articles are publicly available in the publications section of the ATVS group webpage.)
-
[PR2015] J. Galbally, M. Diaz-Cabrera, M. A. Ferrer, M. Gomez-Barrero, A. Morales and J. Fierrez, "On-Line Signature Recognition Through the Combination of Real Dynamic Data and Synthetically Generated Static Data", Pattern Recognition, Vol. 48, pp. 2921-2934, September 2015, [DOI]
-
[PAA2010] J. Fierrez, J. Galbally, J. Ortega-Garcia, M. R. Freire, F. Alonso-Fernandez, D. Ramos, D. T. Toledano, J. Gonzalez-Rodriguez, J. A. Siguenza, J. Garrido-Salas, E. Anguiano, G. Gonzalez-de-Rivera, R. Ribalda, M. Faundez-Zanuy, J. A. Ortega, V. Cardeñoso-Payo, A. Viloria, C. E. Vivaracho, Q. I. Moro, J. J. Igarza, J. Sanchez, I. Hernaez, C. Orrite-Uruñuela, F. Martinez-Contreras and J. J. Gracia-Roche, "BiosecurID: A Multimodal Biometric Database", Pattern Analysis and Applications, Vol. 13, n. 2, pp. 235-246, 2010.
Please remember to reference article [PR2015] on any work made public, whatever the form, based directly or indirectly on any part of the BiosecurID-SONOF DB.
|
|
|
|
|
|
