Content Reconstruction Using Keystroke Dynamics: Preliminary Results

Abstract

Keystroke dynamics is a technique used to verify the identity of a person, by investigating the way a person types on a keyboard. This can be used for example in combination with a password to get access to a computer system. In that case not only the correctness of the password is checked, but also if the manner in which it was typed is correct. This is because the typing rhythm of a person is fairly unique. In this research we investigated the possibility to reconstruct the content of the typed text from the keystroke dynamics typing data. We will show that under highly optimal conditions, a simple English sentence could be reconstructed from timing information only. These results indicate however that under ordinary circumstances, at this moment, it is not likely that an attacker can reconstruct a random text from the keystroke dynamics data. However this is a preliminary study and the specific techniques used in the method we employed can be optimized, which might result in a different conclusion. Our results indicate that an attacker can train the system by using his own keystroke data or similar data of a set of accomplices.