1. Enhancing keystroke dynamics accuracy with optimal SVM kernel usage.
- Author
-
Yaacob, Mohd Noorulfakhri, Idrus, Syed Zulkarnain Syed, and Mohammed, Abdul Hapes
- Subjects
RACIAL classification ,COMPUTER users ,TIME management ,BIOMETRY - Abstract
Behavioral biometric traits are not fully distinguished in recognition tasks, but they can improve the overall performance of biometric recognition systems by adding them. The behavioral biometric studied in this paper is related to keystroke dynamic. This paper examines the touch keystroke dynamics of computer users for the purpose of identifying their culture by using four different SVM kernels. It has been confirmed that racial classifications can be made by gathering keystroke data from 250 respondents representing various culture in Malaysia. Results show that different culture categories display different typing patterns. The classification is made using four SVM kernels and a comparison of the accuracy results is shown. The four kernels are Linear, Quadratic, Cubic and Fine Gaussian. The linear kernel has provided the highest accuracy and consistent readings compared to other kernels for the four features evaluated for dynamic keystrokes, namely press-press time, release-release time, press-release time and release-press time. The linear kernel has the highest accuracy reading of 92.4% for classification using press-press time features for the Malay vs Chinese category. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF