User authentication and distinguishing child users from adults with keystroke dynamics

Uzun, Yasin
Keystroke Dynamics, which is a biometric characteristic that depends on typing style of users, could be a viable alternative or a complementary technique for user authentication if tolerable error rates are achieved. Moreover, biometric data can also be used for inferring personal characteristics. Therefore it is possible to benefit from Keystroke Dynamics to predict infor- mation, such as age and gender. In this thesis study, the performance of artificial neural network algorithms for Keystroke Dynamics based authentication is measured using a publicly available dataset. For this pur- pose, comparative tests of di erent algorithms for training neural networks are conducted and an equal error rate of 7.73 percent with Levenberg-Marquardt backpropagation network is achieved as a result. Regarding to detecting age group and gender information based on typing data, classification accuracies for 13 di erent algorithms is assessed. For this purpose, a new typing dataset from 100 users including male and female, adult and child subjects is collected. For age group detection, average error rates down to 8.2 percent is achieved using k-nearest neighbor algorithm. On the other hand, the minimum error rate recorded for gender prediction was 40 percent, using the same dataset and methodologies that are used for age group detection. The dataset and implementation for the whole experiment and test procedure is made publicly available to promote future works focusing on this subject
Citation Formats
Y. Uzun, “User authentication and distinguishing child users from adults with keystroke dynamics,” Ph.D. - Doctoral Program, 2013.