Sequence Alignment of Dynamic Intervals for Keystroke Dynamics based User Authentication

Due to the widespread use of cloud systems nowadays, strong authentication mechanisms are needed in order to secure personal files. Out of diverse ways to improve secure authentication, keystroke dynamics is of interest because it is inexpensive and needs no extra hardware systems. Furthermore, authentication systems with proper machine learning algorithms can acquire humans' typical typing behaviors from their keystroke dynamics, which entails difficulty for imposters to imitate a legitimate user's typing behavior. In this paper, we introduce sequence alignment algorithm with dynamic interval features (SADI) from keystrokes to model behavior-based authentication system. An interval feature is basically the length of each attribute label and it is used in a sequence alignment algorithm to divide every attribute into sections. However, dynamic interval features, proposed in this research, are similar to interval feature but they divide every attribute into different number of sections. Dynamic interval features are chosen to maximize comparison capability of similarity measures from keystroke data. Experimental results on the CMU public benchmark dataset indicate that the proposed SADI is comparable to and sometimes outperforms other published methods.