Behavior-Based Continuous User Authentication Detection System

Abstract

Continuous authentication offers a vital layer of protection by monitoring user activity beyond simple passwords or fingerprints scans. In this work, we present a system that actively and unobtrusively analyzes how individuals type and use their mouse throughout active desktop sessions. By gathering and processing behavioral data, our approach creates unique user profiles and employes modern machine learning- specifically Random Forest and XGBoost- to distinguish legitimate users from imposters within seconds. Experiments show this technique achieves accuracy above 90% with minimal false alarms, making it a practical solution for preventing sessions hijacking and unauthorized access in both workplaces and at home.