Computer user identification using the mean and the median as features

An effective method for securing access to computers using a password is very important. Real time data was collected over a period of five weeks from six users by measuring time durations between keystrokes entered in a password. A phrase was used as the password. Extracting features from a set of measurements is a common practice in pattern recognition. Two methods are used. One of the methods used is based on dividing the set of measurements into three sets and then computing the mean value of each set (the three means). The other method is using the neighborhood averaging where each measurement is represented by itself and the other two neighbors. Even though the methods save memory, the classification results are not as good as hoped for. These methods do not lend themselves nicely to the problem at hand.