User identification and authentication in browser environments via machine learning

Abstract

Achieving secure and efficient user identification on computer systems necessitates the deployment of strong protective mechanisms, given that conventional password approaches are insufficient to counter significant security threats. Behavioral biometric technologies have been developed to address these security challenges. This study focuses on user authentication via mouse movement dynamics, proposing a novel biometric approach for network administrators who exhibit unique mouse movement patterns. The method leverages mouse movement data over five and ten-second intervals, using features extracted from these data to identify frequent usage areas. Five machine learning algorithms were evaluated, with the Random Forest algorithm demonstrating superior performance. The method achieves a FPR of 0.85% and a FNR of 29.17%, underscoring its potential for enhancing security in network administration tasks. The dataset was generated from mouse movement during training sessions and various competitions, and features were extracted and classified to evaluate the system’s accuracy. The study concludes that Random Forest is the most effective algorithm for this application, meeting regional biometric system standards and suggesting potential for widespread implementation in corporate environments.