Collection and Statistical Analysis of a Fixed-Text Keystroke Dynamics Authentication Data Set

Keystroke dynamics authentication is a promising method of improving account security with minimal detriment for user convenience. While there is an abundance of research, there is a lack of available data sets. In this study, data sets for keystroke dynamics authentication were collected for a set of 6 passwords from a group of participants, and a correlation algorithm was developed to analyze and use these data sets for authentication. The experiments aim to produce data for keystroke dynamics authentication benchmarking, and to show the effect of typing speed and consistency, password length and entropy on prediction accuracy. Through simple correlation methods, the authors achieve an Equal Error Rate varying between a range of 2.57% and 29.7%. These result give insight into what may cause the accuracy to vary depending on the person and the password.