A document classification benchmark dataset on cybersecurity policy domain

Researchers from KAIST developed the first benchmark dataset for classifying cybersecurity policy documents, containing 1,723 expert-labeled samples across Strategy, Legislation, and Guideline categories. Using BERT-based models, they achieved near-perfect classification accuracy, with XLM-RoBERTa-base reaching 99.99%. The dataset and code are publicly released to support further research in cybersecurity policy analysis.