RHU KeyStroke Dynamics Benchmark Dataset

Touch Screen Phone based keystroke dynamics dataset



This webpage presents a new keystroke dynamics dataset. It is a supplement of the paper “RHU Keystroke: A Mobile-based Benchmark for Keystroke Dynamics Systems”, by Mohamad El-Abed, Mostafa Dafer and Ramzi El Khayat, published in IEEE ICCST 2014 conference that can be found here: http://ieeexplore.ieee.org/document/6986984/ It is the first public dataset collected using a touch screen phone and full soft-keyboard characters. KeyStroke (Medium)

Protocol and Population

All participants have participated in 3 sessions. During the first session, the individuals were able to train themselves on the typing of the password on the keyboard as long as they wanted, because it is not their usual password and does not have a preexisting typing habit or pattern for it.

For the following sessions, they were authorized to train just a little bit because touch smartphones’ operating systems have different keyboards, and we wanted them to get used to the different keyboard.

  • –51 individuals
  • –Password: rhu.university
  • –15-20 trials, 3 sessions
  • –3-30 days separating sessions
  • –985 acquisitions

51 individuals have participated in the acquistion process by typing the password “rhu.university” 15 times spread on 3 sessions with time period between 3 to 30 days separating each session with some of them having more than 5 acquisitions on the last session; on average, the separation period between one session and another is 5 days.

We have 985 available acquistions, and the average number of acquisitions per user is 17 with 40 of them having more than 17 acquistions.


Db_font (Medium) Gender and age are chosen by the user during account creation. Gender: 0: Female 1: Male0 Age: 0: between 7 and 18 years old 1: between 1 and 29 years old 2: between 30 and 65 years old The Password is common among all users: rhu.university You can download the dataset  RHU KeyStroke Dynamics Dataset Download (3634 downloads) . (.xls)
The data are stored in an SQLite file (keystroke.db) in order to be easily processed. (Soon)

Collected Data

New user, gender and age Training mode Extracted features:
  • key pressure and a key release
  • key release and a key pressure
  • two key pressures
  • two key releases
  • date of each trial
  • overall time for non-multi-touch typing users
  • Typing correction not allowed

Please, cite the database with this reference :
               title = {RHU Keystroke: A Mobile-based Benchmark for Keystroke Dynamics Systems},
               author = {El-Abed, Mohamad and Dafer, Mostafa and El Khayat, Ramzi},
               Booktitle = {Proceedings of the 48th IEEE International Carnahan Conference on Security Technology}
               year = {2014}   

PR Feature Analysis

Mostafa, Physical Keyboard, pr (Medium)
the extracted PR Feature using a physical keyboard
Mostafa, Touch Keyboard,pr (Medium)
the extracted PR Feature using a touch-screen soft keyboard
Mostafa, Touch Keyboard (Medium)
 the extracted PR Feature along with all extracted features (PP, PR, RP, RR)

For the PR feature (bottom right), this figure shows that the difference between the collected timings of all password’s characters is not significant for the same acquisition. We can relate this insignificance due to the fact that the PR feature is the timing of the finger while touching the screen. So since the touch screen is a flat surface, all the soft buttons take almost the same time to press and release. Whereas on a physical keyboard, a button is physically pressed and released, thus the timing is different among the buttons.

Hardware used to collect benchmark

nokia phone (Medium) Nokia Lumia 920 Capacitive Screen: 4.5” Multi-Touch 768×1280 (332 ppi) Weight: 185 g
A newer published dataset can be found here: Multi-Orientation Benchmark (755 downloads) A data visualizer may be found here: https://www.coolestech.com/keystroke-web-visualizer/

Leave a comment

Like Us!

Subscribe to newsletter!

Enter Your Email: