Recognition systems based on time durations between terminal keystrokes
S.A. Bleha, Charles R. Slivinsky · 1988
This research describes a new pattern recognition approach to securing access to computer systems. By performing real-time measurements of the time durations between the keystrokes when a password is entered and using pattern recognition algorithms, three online recognition systems were devised and tested. The study was partitioned into two phases. Phase I provided the guidance to parameterize and test three recognition systems built in Phase II. Among the most important parameters considered in Phase I are the password length, the type of password, the threshold, and the classifier most appropriate to use for recognition. The use of Fisher's linear discriminant for dimensionality-reduction followed by the minimum-distance classifier produced an error of 1.1%. However, the use of two entries of the password and a new shuffling technique produced lower errors. In the three systems of Phase II both the minimum-distance classifier and the Bayes classifier were implemented. Two types of passwords were considered: phrases and individual names. A fixed phrase was used in the identification system. Individual names were used as passwords in the verification system and in the overall recognition system. All three systems were tested and evaluated. The identification system used ten volunteers and gave an indecision error of 1.17%. The verification system used twenty-six volunteers and gave an error of 8.1% of rejecting valid users, and an error of 2.8% of accepting invalid users. The overall recognition system used thirty-two volunteers, and gave an error of 3.1% of rejecting valid users and an error of 0.5% of accepting invalid users.