Computer users verification using the perceptron algorithm

The perceptron algorithm was used to verify the identity of computer users. By performing the real-time measurement of the time durations between the keystrokes entered in the user's password, data was collected over a period of eight weeks from 10 valid users and 14 invalid users. The password used was the user's name. Decision functions were derived using part of the data (training data) to compute the weight vectors. The decision functions were then applied to the remaining data (testing data) to verify the users. An error of 9% in rejecting valid users, and an error of 8% in accepting invalid users were achieved. >