Behavioral Biometrics for Continuous User Authentication in Mobile Devices

Abstract

The increasing reliance on smartphones for financial, healthcare, professional, and personal services has intensified the need for secure and unobtrusive user authentication beyond initial login. Behavioral biometrics offer a promising basis for continuous authentication by exploiting distinctive interaction patterns such as touch dynamics, keystroke behavior, gait, device motion, sensor responses, application usage, and contextual information. This comprehensive review critically examines recent developments in behavioral biometric authentication for mobile devices, with emphasis on machine-learning and deep-learning techniques, multimodal fusion, adaptive verification, and practical deployment considerations. The reviewed literature demonstrates that conventional machine-learning approaches remain effective for structured behavioral features, while deep-learning models provide stronger capabilities for learning complex temporal and multimodal representations. Multimodal systems can improve robustness by integrating complementary behavioral signals, although their practical adoption is constrained by computational cost, battery consumption, privacy risks, sensor variability, and behavioral drift. The review also highlights persistent challenges involving non-standardized datasets, small user populations, inconsistent evaluation protocols, cross-session and cross-device generalization, spoofing, adversarial attacks, and limited real-world validation. Future research should prioritize privacy-preserving federated learning, lightweight and explainable models, adaptive behavioral profiling, edge intelligence, context-aware authentication, and robust multimodal systems to support trustworthy and scalable continuous mobile authentication.