iSignDB: A database for smartphone signature biometrics

Abstract The signature has long been in use for the user verification. These signatures have user specific features that differentiate the individual for authentication. The signature verification can be offline or online. The offline verification considers only the static features of the signatures through the signature image, while the online verification considers various dynamic features associated with the signature such as pen pressure, pen tilt angle, velocity, acceleration, pen up and pen down, etc at various time stamps which are recorded using special digitizing tablets such as Wacom devices (STU-500, STU-530 and DTU-1031) [ 2 , 15 ] etc. In todays scenario, smartphones are widely used world-wide, and come equipped with various sensors e.g. accelerometer, gyroscope, magnetometer, GPS, etc. able to capture sensor logs and as well used in the literature to capture the dynamics of users’ behaviour while a signer signs on his smartphone. However, there is scarcity in the publicly available databases for the online signatures collected using smartphone. In the present work, we describe biometric signature dataset iSignDB captured using smartphone. The iSignDB [ 7 , 11 ] consists of the genuine signature samples of a user as well as the skilled forgery samples where imposter was given multiple attempts to mimic the mannerisms of the original signer before giving skilled forgery samples. A total of 30 samples towards the genuine signature over 3 sessions with 10 samples per session while 15 samples of the skilled forgery with 5 samples per session were collected. Each of the session were at least 15 days apart. The iOS and Android based smartphones (namely iPhone7 and Redmi Note 7) were used for the data collection. The sensors used to collect this data, present in the smartphone are the gyroscope, magnetometer, GPS, and accelerometer. Smartphones having sensors any one lesser than these four, were not used for data collection, in order to have consistent number of features under each sample. They generate the following sensor readings: angular velocity, acceleration, orientation, geomagnetic field in the x, y, and z directions, position, which is collected in a licensed MathWorks cloud account using the MATLAB Mobile App installed in the smartphone, the form of a multitude of sensor logs. Each sample has image of the signature along with sensor readings. Some of the publicly available smartphone biometric signature databases are DooDB [3] , MOBISIG [4] , eBioSign DS 2 [8] , etc. in which at least acceleration sensor reading is present but the iSignDB ensures these five of the sensor readings (acceleration, angular velocity, magnetic field, orientation, position) under each sample. This dataset can be successfully used to design smartphone biometric signature authentication system which is robust against a number of spoof attacks [12] , [13] , [14] , [15] . As every user has a unique way of handling his/her smartphone which varies over different level of emotional intelligence of the user over a time period, this dataset can also be used for behavioural analysis of the users.