Dynamic shuffling was also evaluated as a process applied to training samples for neural networks as a means of enhancing sample classification and reducing false acceptance and rejection rates during keystroke analysis

While most previous keystroke biometric studies dealt with short input like passwords, we focused on long-text input for applications such as identifying perpetrators of inappropriate e-mail or fraudulent Internet activity. A Java applet collected raw keystroke data over the Internet, appropriate long-text-input features were extracted, and a pattern classifier made identification decisions. Experiments focused on the system’s usability under ideal conditions: copy task, long-text input (600 characters), same keyboard for enrollment and testing, and subjects aware of the nature of the study and instructed to type naturally. Essentially 100% identification accuracy was achieved on 8 subjects typing the same text. This accuracy decreased in going to 30 subjects, on copying different testing texts, and on progressively reducing the length of the testing text. In summary, we found the keystroke biometric effective for identifying up to 30 users inputting text under the following conditions: sufficient training and testing text length, sufficient number of enrollment samples, and same keyboard type used for enrollment and testing.