Keyboard User Test Of The Weighted Principal Component Classifier,

The traditional verification of the access to a computer system is the single password. The drawback of the password is that it is easily given away. Therefore there are great security threats in the access to information and resource in computer, hi an effort to confront the threats biometrics (such as fingerprints, keystroke dynamics, and so on) combined with the traditional verification methods may greatly increase the system security. In view of the fact that it is not necessary for the verification of keystroke dynamics to utilize the additional equipment and the realization is relatively cheap, in this thesis we study the verification of users by means of the combination of keystoke dynamics and password. Based on the previous study, we adopt SVM (Support Vector Machines) to recognize the individual keystroke pattern. Through experiments we gain better results.In addition, based on PCC (Principle Component Classifier), we study a kind of kernel-based classifier. Adopting the proper weighted functions, we obtain WPCC (Weighted PCC) which has better generalization and ability to deal with the outliers than PCC. Analysis and experiments testify the superiority of WPCC in handling outliers and generalization.