A Novel Hybrid Threat Modeling Framework for IoT Security using STRIDE-DREAD and Machine Learning

Abstract

The rapid growth of Internet of Things (IoT) deployments has increased security risks due to diverse device vulnerabilities, large scale interconnected environments, and the heterogeneity of communication protocols. Traditional threat assessment methods such as STRIDE and DREAD provide a structured foundation for identifying and categorizing security risks, yet they lack automated, real-time detection capabilities required for modern IoT systems that operate in dynamic and resource-constrained environments. To address these limitations, this study presents a hybrid threat modeling framework that integrates machine learning with STRIDE–DREAD to enhance threat identification, prioritization, and quantitative risk analysis. An ML-based classifier is trained on the CIC-BCCC-NRC TabularIoTAttack-2024 dataset to detect and categorize various IoT attack types, with particular emphasis on DDoS variants due to their high prevalence. Ensemble learning techniques are applied to pre-processed network traffic, enabling accurate, scalable, and computationally efficient classification suitable for deployment on lightweight IoT hardware. The proposed system achieves 92.5% detection accuracy, surpassing conventional STRIDE–DREAD assessments by 10–15% while providing enriched decision support for security analysts. Overall, the results demonstrate that integrating ML with established threat modeling methods significantly improves automation, reduces manual evaluation time, and strengthens the precision, adaptability, and operational reliability of IoT security assessment frameworks.

Keywords