User keystroke authentication based on stable digraph pairs

Both Joyce-Gupta method and Bergadano-Guntetti-Picardi method are well-known for user keystroke dynamics authentication by comparing latency of two consecutive keystroke, termed digraph, of test sample with that of reference sample. By millisecond units, one does not always type keys with the same keystroke latency. However, it seems that some digraphs are always faster than other slow digraphs, which is stable and unique to person. In this paper, we call such fast and slow digraph pair a stable digraph pair, based on which we propose a new method for user keystroke authentication. By combining the previous methods, we have tested our approach on 7 individuals, for a total of 35 samples achieving both false alarm rate and impostor pass rate of 0%, which is better result than those methods alone.