Towards Security Awareness Enhancement using Dynamic and Adaptive Behavior Learning Models
Preshika Basnet, Isha Nepal, Rasib Hassan Khan · 2025
As cyber threats continue to evolve alongside various forms of technological advancements, human behavior remains a critical vulnerability in modern security infrastructures. Understanding and addressing users’ personalized behavioral traits is vital to strengthening an organization’s overall security posture, as even well-designed technical defenses can be compromised by poor decision-making or lack of awareness. This paper proposes a dynamic, AI-driven cybersecurity framework that leverages User and Entity Behavior Analytics (UEBA), Machine Learning (ML), and Large Language Models (LLMs) to deliver personalized, real-time security awareness and training modules. Built on a feedback- and feedforward-enhanced Input-Process-Output (IPO) model, the system analyzes multi-modal behavioral data to detect anomalies, optimize policies, and adapt training content dynamically. The framework enhances engagement, improves long-term security behavior, and strengthens organizational resilience against emerging threats. Empirical evaluation demonstrates the model’s scalability and effectiveness, while addressing challenges related to data privacy, integration complexity, and contextual adaptability.