Fog-Based Transformer Learning for Multi-Class Classification of RPL-Based Attacks

Identifying attacks on Routing Protocol for LowPower and Lossy Networks (RPL) is quite challenging in resource-limited environments. This paper develops a transformer learning system implemented through fog infrastructure that classifies multiple RPL-based attacks. The evaluation process includes extensive simulations and dataset generation to reproduce real attack behaviors and varied network scenarios. The proposed model applies deep learning approaches, SMOTE and weighted sampling to balance data for better detection outcomes. The experimental results indicate excellent classification performance with an accuracy of 97.17 %, AUC-ROC scores of $\mathbf{9 9. 7 6 \%}$, and MCC scores of $\mathbf{9 1. 8 0 \%}$. The research establishes transformer-based models as robust solutions for resource-limited fog environments that produce efficient intrusion detection systems for Internet of Things (IoT) networks.