The document presents an approach to enhance the security of the Border Gateway Protocol (BGP) by addressing issues related to dynamic network behavior and unbalanced datasets that hinder attack detection. It introduces a method that converts binary classification problems into multiclass classification, utilizing affinity propagation to balance the datasets, and trains an Extreme Learning Machine (ELM) for improved performance. Experimental results show significant increases in F1 scores, particularly with specific feature selection techniques, outperforming state-of-the-art methodologies.