Keystroke Dynamics Authentication with 3D Trajectory Dissimilarity
Tanapat Anusas-amornkul, Thagoon Phigoonkhao · WSEAS TRANSACTIONS ON SYSTEMS AND CONTROL · 2025
Password authentication is a well-known and widespread technique for user authentication. However, with highly developed technologies nowadays, password authentication has become less effective in guarding against a password-guessing attack. Therefore, two-factor authentication is used to enhance password authentication. In this work, the password and keystroke dynamics of a username are used to enhance the classical authentication scheme, which is known as two-factor authentication. Keystroke dynamics is a biometric authentication involved with individual typing behavior parameters: hold time, interkey time, and latency time. This work proposed extracting basic statistical features, which are mean and median, to create 3D trajectory curves. The 3D curve of a user is compared point-by-point with a master keystroke profile by computing using Euclidean, Manhattan, and Mahalanobis distances. The result demonstrates that a 3D trajectory dissimilarity using mean and Euclidean distance methods gives the highest efficiency using the Confusion matrix with accuracy and F1-Score of 99.07% and 0.95, respectively. In addition, a false acceptance rate (FAR), a false rejection rate (FRR), and an equal error rate (EER) for the proposed work are 0.74%, 2.66%, and 0.93%, accordingly.