A comparison of keystroke dynamics techniques for user authentication

Recently, in the Internet, one of the most security vulnerabilities is a weak password setting. There are several ways to make the password harder to guess by increasing the number of characters, password complexity, or changing the password more often. However, using only passwords for authentication may not be enough because passwords can be written down or exposed to others easily. Therefore, several researchers are solving this problem by adding keystroke dynamics to a username or a password to strengthen the authentication process. In this work, three keystroke dynamics techniques, i.e. statistics using confidence interval, k-means clustering, and trajectory dissimilarity, are implemented and compared with the same dataset. The performance metric is accuracy. In addition, pseudocodes for the techniques are also presented. From the experiment, the trajectory dissimilarity technique gives the best accuracy at 96% among others.