Continuous Biometric User Authentication Based on Keystroke Dynamics

Abstract

The article examines modern approaches to biometric user authentication, focusing on behavioral methods, particularly keystroke dynamics (Keystroke Dynamics, KD). It outlines the advantages and limitations of physiological and behavioral biometrics and substantiates the feasibility of using keystroke dynamics as a means of continuous background authentication. The study includes an analysis of both classical and recent research in this field, including works employing ensemble algorithms and deep learning techniques. A modular architecture of a continuous authentication system is proposed, integrating data collection and processing, classification, and adaptive user profile updating. Key accuracy metrics (FAR, FRR, EER, TTD) are discussed, as well as the potential integration of keystroke dynamics with other authentication factors in multi-factor systems. The analysis of keystroke dynamics enhances the security level of information systems without compromising user convenience and offers potential applications in military, financial, and corporate environments