K. Barrelle et al., USER VERIFICATION THROUGH POINTING CHARACTERISTICS - AN EXPLORATION EXAMINATION, International journal of human-computer studies, 45(1), 1996, pp. 47-57
Citations number
28
Categorie Soggetti
Psychology,Ergonomics,"Computer Sciences","Controlo Theory & Cybernetics","Computer Science Cybernetics
Previous research has highlighted that all data security systems rest
on the assumption of accurate user verification. Although much researc
h has focused on speech and keystroke characteristics, no study has ex
amined the potential of indirect cursor control devices, such as mouse
or pen, as a means of personal user verification. This paper presents
the result of an exploratory examination of the measurement of indivi
dual users' behaviour elicited while using two indirect input devices
(puck and pen). Ten subjects undertook two series of 500 trials using
the puck and pen in a counterbalanced design. Each trial involved guid
ing the cursor to one of ten possible targets, projected upon one of f
ive angles of projection and one of two distances. Data recorded compr
ised a series of (x,y) coordinates and associated time stamps of each
component of the movement. Five parameters for each full movement were
extracted--total time to reach the target, maximum velocity of the mo
vement, maximum acceleration for the movement, time to maximum velocit
y of movement, time to maximum acceleration of the movement. Results o
f a series of Multivariate Analyses of Variance indicated that for bot
h the puck and the pen data, there were statistically significant diff
erences between subjects for each of the five parameters. This set of
discriminating parameters was then embedded in a verification paradigm
. Results indicated that the average error rates varied between 39% (S
.D. = 2%) and 14% (S.D. = 2%) for the puck and 38% (S.D. = 2%) and 12%
(S.D. = 2%) for the pen. Error rates were dependent upon the number o
f samples used when making the verification decision. This research su
ggests that pointing device measurements using the current parameters
are not a viable method of user verification by themselves. However, w
ith average error rates as low as 12%, it may be reasonable to integra
te them into a multi-modal security system. Practical implications and
future research directions are discussed. (C) 1996 Academic Press Lim
ited