Touch-based Static Authentication Using a Virtual Grid

摘要

Keystroke dynamics is a subfield of computer security in which the cadence of the typist's keystrokes are used to determine authenticity. The static variety of keystroke dynamics uses typing patterns observed during the typing of a password or passphrase. This paper presents a technique for static authentication on mobile tablet devices using neural networks for analysis of keystroke metrics. Metrics used in the analysis of typing are monographs, digraphs, and trigraphs. Monographs as we define them consist of the time between the press and release of a single key, coupled with the discretized x-y location of the keystroke on the tablet. A digraph is the duration between the presses of two consecutively pressed keys, and a trigraph is the duration between the press of a key and the press of a key two keys later. Our technique combines the analysis of monographs, digraphs, and trigraphs to produce a confidence measure. Our best equal error rate for distinguishing users from impostors is 9.3% for text typing, and 9.0% for a custom experiment setup that is discussed in detail in the paper.