Enhanced Keystroke Vector Dissimilarity using Machine Learning Techniques

Nowadays, a computer is an important device for accessing the digital world. When a user logs into a system, he has to provide his credentials to the system using a typical password, but this scheme is not secure anymore if the password is leaked or stolen. Therefore, keystroke dynamics can be used as a biometric authentication to increase the security of the system without adding a new device. The objective of this work is to improve the performance of a keystroke vector dissimilarity authentication by using machine learning techniques. Four machine learning models, which were logistic regression, support vector machine, decision tree, and random forest models, were studied. In addition, the number of input features, one and three features, was under-investigated to find the best performance. The results indicated that the best accuracy was 98.49% using 3 input features with a decision tree model, compared with 96.97% accuracy in a previous work.