Today, many organizations and companies employ biometric authentication technologies in addition to traditional username/password methods to secure their systems and ensure safety. Biometric systems are widely used two-factor authentication systems that provide high security in identity verification processes by utilizing unique biological features. However, there are certain challenges associated with relying solely on biometric characteristics. Factors such as false acceptance or rejection rates, vulnerability to spoofing attacks, and the need for specialized hardware can affect these systems. Therefore, incorporating machine learning methods in the field of behavioral biometrics has the potential to enhance accuracy rates. This study examines a behavioral-based authentication method on mobile devices. The objective is to develop a secure authentication mechanism by utilizing data obtained from users' keystrokes. In this scope, keystroke data from 47 employees with diverse characteristics were collected from their mobile devices. Training and testing processes were performed using machine learning algorithms, followed by performance evaluation. The analysis revealed that a tree-based algorithm called Random Forest Classifier achieved the highest accuracy rate (97%). This outcome demonstrates the effective utilization of behavior-based biometric authentication methods. The findings also indicate that data obtained from users with different characteristics contribute to better predictions by the model and enhance the security level. ORCID NO: 0000-0003-1786-4617, 0000-0003-0604-2749, 0000-0003-4667-1929, 0009-0001-6497-4475
Anahtar Kelimeler: keystroke dynamics, behavioral biometrics, two factor authentication, machine learning
|
