Adaptive Cyber Risk Scoring Models for Digital Learning Environments

Digital Learning Environments (DLEs) such as Learning Management Systems (LMS), Massive Open Online Courses (MOOCs) and virtual classroom platforms have become an essential piece of education infrastructure around the world. Today, the education industry has seen more than 1,500 confirmed data breaches reported every year, and around 2,300 cyberattacks reported every week.But current static risk assessment models like CVSS v4.0 or NIST SP800-30 do not account for the temporal, behavioral and academic-contextual characteristics that distinguish DLE threat landscapes. In this paper, a new model, ACRSM-DLE (Adaptive Cyber Risk Scoring Model for Digital Learning Environments), that dynamically scores the level of risk in digital learning environments based on the integration of infrastructure vulnerabilities, user behavioral analysis, real-time threat intelligence, and a new Academic Context Risk Factor (ACRF) is presented. A new hybrid machine learning architecture is used in the model, which includes Random Forest, XGBoost, LSTM networks and Isolation Forest, Bayesian score fusion and Adaptive Risk Threshold Recalibration (ARTR) mechanism. Experimental assessment in four institutional DLE deployments, with 187,000+ learners, shows a 41.3% false-positive rate reduction, a 34.7% improvement in detection precision, an F1-Score of 0.939, an AUC-ROC of 0.974, and a mean score update latency of 287 ms, all significantly better than static baselines. The model complies with the requirements of FERPA, GDPR Article 32 and ISO/IEC 27005: 2022.

Hali tarjima qilinmagan