New RoBERTa Framework Tackles Class Imbalance in Cybersecurity Vulnerability Classification

A new research paper proposes a Hierarchy-Aware RoBERTa framework to address class imbalance in cybersecurity vulnerability classification using the Common Weakness Enumeration (CWE) taxonomy. The framework explicitly incorporates CWE structural information through learnable parent-class embeddings to maintain taxonomic consistency. Experiments on a CWE Research Concept dataset showed that this approach achieved a weighted F1-score of 0.76, outperforming traditional oversampling techniques like SMOTE and ADASYN, and significantly improving performance on minority classes. AI

IMPACT This research offers a more principled approach to handling imbalanced datasets in specialized domains like cybersecurity, potentially improving the accuracy of vulnerability detection systems.

RANK_REASON The cluster contains an academic paper detailing a new methodology for AI model training. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• ADASYN: Adaptive synthetic sampling approach for imbalanced learning
• Bert
• Common Weakness Enumeration
• CWE Research Concept dataset
• Hierarchy-Aware RoBERTa
• Smote

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →