Investigation of keyboard digraphs informational parameters for keystroke-based identification tasks of computer networks users

Abstract

The accuracy of keystroke based authentication methods for computer networks users is determined by a number of factors, the main of which are: user classification algorithm, number of experiment participants with different keystroke experience, method of data entry and hardware platform on which authentication systems testing is performed. Another very important factor, namely, the kind of timing features of keystorke, on the basis of which user profiles are built, is very often not taken into account.

In the first part of the work the different timing features of keystroke are analyzed. Three main classes are identified: timing features of keyboard monographs, timing features of consecutive keyboard events and integral features of typing. Given the accuracy of authentication and the required volume of training and authentication samples, 4 timing features of keyboard monographs (2 absolute and 2 relative) and 15 timing features of keyboard digraphs (3 absolute and 12 relative) were selected as experimental informative parameters of keystroke.

In the second part of the work, the most informative timing features of the keystroke monographs and digraphs have been found using "Keystroke Dynamics Benchmark Data Set " and Orange software. The assumption of different informativeness of the keystroke monographs and digraphs timing features was experimentally confirmed. The most useful (provides the highest authentication accuracy) timing features of keystroke dynamics is relative features of keystroke time parameters, which contains a flight time.