103
User Authentication
accurate results than texts containing only digits [14]. Bleha et al. [1] showed that longer
words minimize the misclassification rate.
Many type of keystroke features have been used in literature. A percentage distribution of
keystroke features has been indicated in Figure 6.3, where combination of key hold time of a
single key and latency times between two subsequent entered characters are the mostly used
features. Some researchers say that di-graph time is the most elementary timing feature in
keystroke dynamics, but only 12% of di-graph timing features have been used. Key pressure
is also applied as an effective feature in keystroke dynamics where a special keyboard or a
pressure-sensitive keyboard is required. As  technology evaluation grows, many advance
and sensitive hardwares have been used in smart phones or portable devices. Through the
advance sensing devices, we can get some additional effective features such as finger tips
size, velocity of finger movements in different directions [15].
Paragraph
3%
Digits
19%
Password
15%
Long text
31%
Short text
32%
FIGURE 6.2
Percentage distribution of type of texts used in literature.
Key hold + latency
time
50%
Only
latency
time
11%
Di-graph time
12%
Key pressure
18%
Key stroke sound
2%
Key hold time
2%
Others
5%
FIGURE 6.3
Percentage distribution of keystroke features used in literature.