Abstract: Traditional authentication systems relying on static credentials or fixed biometrics are increasingly vulnerable to credential theft, phishing, and spoofing. Behavioral biometrics such as keystroke dynamics and mouse movements provide a more secure alternative but often lack adaptability and add friction. This paper proposes an AI-driven adaptive authentication system that fuses behavioral biometrics with contextual data including device information, location of login, and date and time of login to compute a dynamic trust score. The system adjusts authentication requirements in real time, providing stronger security while maintaining usability. Experimental analysis and literature review suggest that multimodal behavioral and contextual fusion reduces error rates, improves robustness against spoofing, and provides resilience in real-world deployment scenarios.

Keywords: Adaptive authentication, behavioral biometrics, keystroke dynamics, mouse dynamics, risk scoring, privacy- preserving security, multi-factor authentication, continuous authentication.

Downloads: | DOI: 10.17148/IARJSET.2025.1211031