DESCRIPTION OF DooDB

The DooDB database comprises two subcorpora, each one containing a different modality:

Subcorpus 1: Doodles. Participants were asked to draw with their fingertip a doodle on a handheld device touchscreen (see Fig. 1) that they would use as a graphical password on a regular basis for authentication (e.g. instead of the PIN code). There were no restrictions regarding duration or shape. In most cases, users invented their own doodle at the time of acquisition.

Subcorpus 2: Pseudo-signatures. Participants were also asked to draw with their fingertip a simplified version of their signature, which they would also use as a graphical password on a regular basis. This could be, for example, their initials or part of their signature flourish. The main difference between doodles and this modality is that in this case, the dynamic process to produce the drawing is in general composed of natural and well trained movements.

Acquisition was performed using an HTC Touch HD mobile phone. The device has a resistive touchscreen of 2x3.5 in (ca. 5x8.5 cm). The x and y coordinates of the fingertip position are sampled at discrete time values t at 100Hz when the user presses the screen. The coordinate values represent milli-inches, so x values range between [0,2000] (width) and y values between [0,3500] (height). The time interval between consecutive samples is also stored. Some examples of doodles and pseudo-signatures are shown in Fig.2. The acquisition process was divided in two sessions, separated by an average period of two weeks. The donors were asked to draw with their fingertip on the handset touch screen holding it in their own hand, simulating thus real operating conditions. They were allowed to practice their drawings until they felt comfortable with them. Forgeries have also been captured in this database. To perform forgeries, users had visual access to the doodle or pseudo-signature they had to imitate. The acquisition software replayed the strokes on the screen showing their dynamic properties (e.g. speed). This animation was shown to users up to three times, and then they were allowed to train until they felt confident with their forgery. The usage of the replay software makes possible to produce forgeries with a notable degree of accuracy, as can be observed in Fig.2. During the two sessions, the same protocol was followed for each user and modality: 5 genuine samples, then 5 forgeries, 5 genuine samples, followed by 5 forgeries and finally 5 genuine samples. This separation in blocks of 5 signatures allows analyzing intra-session variability. Consequently, at the end of the two sessions, each user had produced 30 genuine drawings (15 per session) and 20 forgeries. In the first session, user n produced forgeries for users n-1 and n-2, while in the second, forgeries for users n-3 and n-4 were produced.

Figure 1. Doodle acquisition setup.

Figure 2. Examples of doodles (left) and pseudo-signatures (right).

FILES FORMAT

The doodles and pseudo-signatures are stored in text files following the following format:

• COLUMN 1: represents the x coordinate.

• COLUMN 2: represents the y coordinate.

• COLUMN 3: this is a synthetic timestamp, representing the time interval in ms between samples.

FILES NOMENCLATURE

Users are numbered by the following pattern:

• [2000,...,2099] for doodles

• [1000,..., 1099] for pseudo-signatures. User 2XXX in the pseudo-signature dataset corresponds to user 1XXX in the doodle dataset.

For each user, there is a folder named SS1 for session 1 and SS2 for session 2.

The nomenclature followed to name the doodle and pseudo-signature files is as follows: SIGN_TTTB_USXXXX_USYYYY_ZZ.txt

• TTT: is the doodle type; GEN for genuine doodles and FOR for forgeries

• B: is the capture block within the session [1, 2, 3]

• XXXX: is the number of the user performing the doodle[1001, 1002, ... , 1029]

• YYYY: is the number of the owner of the doodle being traced [1001, 1002, ... , 1029]. This is the same user for genuine samples and the number of the user whose doodle is being forged in the case of forgeries.

• ZZ: is the number of the sample within the capture session[1, 2, ... , 25]

REFERENCES

For further information on the database, we refer the reader to the open acces article:

• [ACCESS2013]M. Martinez-Diaz, J. Fierrez and J. Galbally, "The DooDB Graphical Password Database: Data Analysis and Benchmark Results", IEEE ACCESS, Vol. 1, pp. 596-605, 2013 (DOI 10.1109/ACCESS.2013.2281773).

Please remember to reference article [ACCESS2013] on any work made public, whatever the form, based directly or indirectly on any part of the DooDB database.