VLAI: A RoBERTa-Based Model for Automated Vulnerability Severity Classification

TL;DR

VLAI is a transformer-based model built on RoBERTa that automatically classifies software vulnerability severity from text, achieving high accuracy and aiding faster triage.

Contribution

This paper introduces VLAI, a novel transformer-based model fine-tuned on a large vulnerability dataset for automated severity classification.

Findings

Over 82% accuracy in severity prediction

Trained on 600,000 vulnerabilities

Open-source model and dataset available

Abstract

This paper presents VLAI, a transformer-based model that predicts software vulnerability severity levels directly from text descriptions. Built on RoBERTa, VLAI is fine-tuned on over 600,000 real-world vulnerabilities and achieves over 82% accuracy in predicting severity categories, enabling faster and more consistent triage ahead of manual CVSS scoring. The model and dataset are open-source and integrated into the Vulnerability-Lookup service.