Abstract
In the contemporary digital landscape, the rapid proliferation of Artificial Intelligence (AI) and the shift toward remote education have intensified the need for robust security frameworks to protect e-learning platforms. As these platforms store sensitive academic, personal, and financial data, preventing unauthorized access has become a critical priority for educational institutions worldwide. Despite their widespread use, traditional authentication mechanisms, such as static passwords and standard Multi-Factor Authentication (MFA), are increasingly inadequate against sophisticated cyber threats. Modern attackers frequently bypass these static barriers through advanced phishing, credential stuffing, man-in-the-middle (MitM) attacks, and session hijacking. Consequently, there is an urgent need for an authentication paradigm that is not only resilient to evolving threats but also adaptable to the diverse behaviors of legitimate users.
This research proposes the development of an Adaptive AI-Powered Multi-Factor Authentication (AI-PMFA) system integrated with a Real-Time Intrusion Detection and Prevention System (IDPS). Unlike conventional MFA, which requires the same secondary verification regardless of the context, the AI-PMFA model leverages machine learning algorithms to evaluate the risk level of every access request in real-time. The system integrates multiple data streams, including biometric verification, behavioral analytics (such as keystroke dynamics and navigation patterns), and contextual risk factors (such as geographic location, IP reputation, and time-of-access anomalies). By calculating a dynamic risk score, the framework can automatically escalate authentication requirements, requesting additional factors only when suspicious activity is detected, or streamline access for low-risk, verified users.
The AI-PMFA system will be developed and validated using comprehensive open-source authentication datasets and trained using anomaly detection models to distinguish between baseline user behavior and malicious intrusion attempts. To demonstrate its practical utility, the framework will be implemented and tested on a dual-environment setup comprising a web-based e-learning platform and a corresponding mobile application. The empirical evaluation will focus on critical performance metrics, including detection accuracy, false-positive rates, system latency, and overall authentication efficiency. Furthermore, the research will assess the usability-security trade-off to ensure that heightened security does not impede the user experience for students and educators. Ultimately, this work contributes to the advancement of cybersecurity and AI-driven identity management, offering a scalable, intelligent, and user-centric solution for securing the future of online learning.