The MOBIKEY Keystroke Dynamics Password Database

URL: http://www.ms.sapientia.ro/~manyi/mobikey.html

Number of subjects: 54 (49 male, 5 female)
Passwords:
- easy: kicsikutyatarka
- logicalstrong: Kktsf2!2014
- strong: .tie5Roanl
Number of samples/subject: at least 60
Number of sessions: at least 3 sessions/user
Age: 19-26, average: 20.61

First order features	Second order features	Explanation	Password's path
Key hold time for each key (HT) Down-down time for each key (DD) Up-down time for each key (UD) Key press pressure for each key (P) Finger area for each key (FA)	Total distance (TD) Total time (TT) Mean velocity (V) Mean x acceleration (MAX) Mean y acceleration (MAY) Mean z acceleration (MAZ) Mean hold time (MHT) Mean finger area (MFA) Mean pressure (MP)	Total distance (TD): length of the path (in pixels) Total time (TT): time needed to type in the password Mean velocity (V): quotient of TD and TT

All features	Second order features
easy.arff logicalstrong.arff strong.arff	easy_secondorder.arff logicalstrong_secondorder.arff strong_secondorder.arff

EER results for different anomaly detectors and feature sets. The standard deviations are shown in parenthesis.

Method	Features	Easy	Logical strong	Strong
Euclidean	secondorder	0.208 (0.174)	0.149 (0.141)	0.181 (0.145)
Manhattan	secondorder	0.202 (0.169)	0.144 (0.140)	0.169 (0.146)
Mahalanobis	secondorder	0.191 (0.182)	0.154 (0.171)	0.159 (0.159)
Outlier count	secondorder	0.208 (0.147)	0.164 (0.140)	0.178 (0.146)
Kmeans	secondorder	0.177 (0.155)	0.136 (0.132)	0.143 (0.137)
Euclidean	all	0.238 (0.186)	0.183 (0.149)	0.195 (0.163)
Manhattan	all	0.203 (0.183)	0.154 (0.140)	0.167 (0.153)
Mahalanobis	all	0.256 (0.140)	0.193 (0.114)	0.210 (0.137)
Outlier count	all	0.160 (0.140)	0.129 (0.126)	0.143 (0.137)
Kmeans	all	0.173 (0.136)	0.128 (0.097)	0.131 (0.106)

EER results for two-class classifiers and different feature sets. The standard deviations are shown in parenthesis.

Method	Features	Easy	Logical strong	Strong
Bayes Net	secondorder	0.074 (0.046)	0.058 (0.040)	0.067 (0.047)
Knn (IBk)	secondorder	0.056 (0.032)	0.048 (0.026)	0.054 (0.036)
Random Forests	secondorder	0.052 (0.029)	0.045 (0.025)	0.051 (0.032)
Bayes Net	all	0.053 (0.039)	0.046 (0.037)	0.049 (0.038)
Knn (IBk)	all	0.073 (0.036)	0.068 (0.033)	0.071 (0.043)
Random Forests	all	0.032 (0.021)	0.033 (0.025)	0.033 (0.022)

Margit ANTAL, László Zsolt SZABÓ, Lehel NEMES (2015), The influence of password difficulty on keystroke dynamics using touchscreen-based devices, The 5th International Conference on Mathematics and Informatics (MathInfo), 2-4 September 2015, Faculty of Technical and Human Sciences, Sapientia University, Tirgu Mures Romania.
Margit ANTAL, Lehel NEMES (2016), The MOBIKEY Keystroke Dynamics Password Database: Benchmark Results, Software Engineering Perspectives and Application in Intelligent Systems, Volume 465 of the series Advances in Intelligent Systems and Computing pp 35-46. [Link]

The research and publication has been supported by the European Union and Hungary and co-financed by the European Social Fund through the project TAMOP4.2.2.C-11/1/KONV-2012-0004-National Research Center for Development and Market Introduction of Advanced Information and Communication Technologies.
We would also like to thank all the students who have participated in the creation of the keystroke dynamics database.