Continuous Authentication by Keystroke Dynamics Using Committee Machines

Abstract

Current authentication mechanisms have a barely addressed problem: they only authenticate a user at the login procedure. If a user leaves the desk without logging out or locking computer session, an intruder has an occasion to use the system.

This paper proposes an authentication methodology that is both inexpensive and non-intrusive and authenticates users continuously while using a computer keyboard. The proposed methodology uses neural networks committee machines to recognize user’s typing pattern, which is a biometric behavioral characteristic. The continuous authentication prevents potential attacks when user leaves the desk without logging out or locking computer session. Some experiments are done to evaluate and calibrate the authentication committee.