Abstract/Summary

This research aims to overcome the drawback of the large amount of training data required for free-text keystroke dynamics authentication. A new key-pairing method, which is based on the keyboard’s key-layout, has been suggested to achieve that. The method extracts several timing features from specific key-pairs. The level of similarity between a user’s profile data and his or her test data is then used to decide whether the test data was provided by the genuine user. The key-pairing technique was developed to use the smallest amount of training data in the best way possible which reduces the requirement for typing long text in the training stage. In addition, non-conventional features were also defined and extracted from the input stream typed by the user in order to understand more of the users typing behaviours. This helps the system to assemble a better idea about the user’s identity from the smallest amount of training data. Non-conventional features compute the average of users performing certain actions when typing a whole piece of text. Results were obtained from the tests conducted on each of the key-pair timing features and the non-conventional features, separately. An FAR of 0.013, 0.0104 and an FRR of 0.384, 0.25 were produced by the timing features and non-conventional features, respectively. Moreover, the fusion of these two feature sets was utilized to enhance the error rates. The feature-level fusion thrived to reduce the error rates to an FAR of 0.00896 and an FRR of 0.215 whilst decision-level fusion succeeded in achieving zero FAR and FRR. In addition, keystroke dynamics research suffers from the fact that almost all text included in the studies is typed in English. Nevertheless, the key-pairing method has the advantage of being language-independent. This allows for it to be applied on text typed in other languages. In this research, the key-pairing method was applied to text in Arabic. The results produced from the test conducted on Arabic text were similar to those produced from English text. This proves the applicability of the key-pairing method on a language other than English even if that language has a completely different alphabet and characteristics. Moreover, experimenting with texts in English and Arabic produced results showing a direct relation between the users’ familiarity with the language and the performance of the authentication system.