This page exists for backup purpose. Official content does not exist anymore; an archive is availabe here https://web.archive.org/web/20130601142020/http://www.epaymentbiometrics.ensicaen.fr/index.php/app/resources/66.


Presentation

Keystroke Dynamics is a biometric modality allowing to recognize users on the way of how they type a password. In this page, you can find a database in the SQLite format to create algorithms and improve the performance of biometric systems based on this modality.

The main objective of this project is to provide a reference database in order to be used by other researchers to test their algorithms. Other modalities have their own databases, and there is a big lack for the keystroke dynamics modality. It seems impossible to really compare the different publications due to the differences of experimental protocols (settings of algorithms, number of enroled vectors…). Data Collection

The data are stored in an SQLite file (keystroke.db) in order to be easily processed.

When enrolling, the following data are saved:

  • the time to type the password

  • the password string

  • the number of enrol error

  • the raw data:

-    Press or Release event
-    Code of the key
-    timestamp of the event
  • extracted features:
-    difference between two press events
-    difference between two release events
-    difference between one press and one release events
-    difference between one release and one press events
-    the concatenation of the previous vectors (V)

Pay attention to the extracted features, the RP event may not be equal to the duration of the key press. It is just the difference of time of following release and press, and not the release and press of the key. KeyStroke Database

The keystroke database must be used only in a scholar research environment. You can download it here

Please, cite the database with this reference :

@INPROCEEDINGS{
	giot2009benchmark,
	author = {Giot, Romain and El-Abed, Mohamad and Rosenberger Christophe},
	title = {GREYC Keystroke: a Benchmark for Keystroke Dynamics Biometric Systems},
	booktitle = {IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS 2009)},
	year = {2009},
	publisher = {IEEE Computer Society},
	address = {Washington, District of Columbia, USA},
}

We have created this database with the intention to test the evolution of the biometric model depending of the keyboard and to have a quite huge number of patterns. Too many studies have been done with very few people and captures.

The following protocol was used. Users were asked to enrol themselves on the system during 1 month with 1 or 2 sessions per week. During the first session, users could learn and test the password greyc laboratory as they wanted. When they felt ready, the capture begins. During the capture process, the users typed six times the password on each keyboard by alternating them (i.e. keyboard 1, keyboard 2, keyboard 1, and so on).

Duration of collection 2 months
Nb of users 133
Nb of users with at least 5 sessions 100
Mean number of captures per user 51
Nb of captures 7555

References

  • R. Giot, B. Hemery, and C. Rosenberger, “Low Cost and Usable Multimodal Biometric System Based on Keystroke Dynamicsand 2D Face Recognition,” in Proc. IAPR International Conference on Pattern Recognition (ICPR), Istanbul, Turkey, 2010.
  • R. Giot, M. El-Abed, C. Rosenberger, “GREYC Keystroke: a Benchmark for Keystroke Dynamics Biometric Systems”, IEEE Third International Conference on Biometrics: Theory, Applications and Systems (BTAS), Washington DC USA, Sept. 28-30, 2009.
  • R. Giot, M. El-Abed, C. Rosenberger, “Keystroke Dynamics With Low Constraints SVM Based Passphrase Enrollment”, IEEE Third International Conference on Biometrics: Theory, Applications and Systems (BTAS), Washington DC USA, Sept. 28-30, 2009.
  • R. Giot, M. El-Abed, C. Rosenberger, “Keystroke Dynamics Authentication For Collaborative Systems”, The IEEE International Symposium on Collaborative Technologies and Systems (CTS), 2009.