moro72842/

cybersecurity-sft-dataset

A curated dataset for training cybersecurity-focused code models with structured JSON output capability.

	Dataset Composition




	

Source Count Percentage Description

CVE Records 10,000 50.0% Multi-turn CVE vulnerability analysis

OpenCodeReasoning (NVIDIA) 5,000 25.0% Chain-of-thought code reasoning

Code-Feedback 5,000 25.0% Multi-turn code debugging and refinement

Synthetic Security (JSON) 5 <0.1% JSON-structured CVE, MITRE ATT&CK… See the full description on the dataset page: https://huggingface.co/datasets/moro72842/cybersecurity-sft-dataset.