Voidreaper2026/cybersec-master-dataset

Cybersecurity Master Instruction Dataset

Overview

A large-scale cybersecurity instruction-tuning dataset in ShareGPT conversational format, assembled from multiple authoritative open sources and deduplicated.

At 1,807,941 deduplicated records, this appears to be one of the larger cybersecurity LLM fine-tuning / instruction-style datasets on Hugging Face, and likely among the larger broad vulnerability-intelligence corpora in conversational/instruction format. It is substantially larger than several known cybersec instruction datasets publicly available at time of publication — however, we make no absolute claim of being the largest without a full audit of all available datasets.

Sources

Source	Records	Description
NVD (National Vulnerability Database)	500,935	Full CVE database back to 2002
OSV (Open Source Vulnerabilities)	754,273	Multi-ecosystem vulnerability database
GitHub Advisory Database	328,525	Security advisories across all ecosystems
Cybersec Causal Reasoning	99,870	System/user/assistant cybersec reasoning triples
Security Stack Exchange	55,930	Real-world Q&A from security.stackexchange.com
ExploitDB	46,457	Public exploit database
MITRE ATT&CK	2,205	Techniques, mitigations, groups, malware (Enterprise/Mobile/ICS)
CISA KEV	1,587	Known Exploited Vulnerabilities catalogue
Kali Linux Tools	790	Tool descriptions, flags, use cases
Vulners	87,063	Vulnerability intelligence including exploits, advisories and CVEs
Total (deduplicated)	1,807,941

Format

All records are in ShareGPT conversational format:

{
  "conversations": [
    {"from": "human", "value": "What is CVE-2021-44228 (Log4Shell)?"},
    {"from": "gpt", "value": "**Summary:** ..."}
  ]
}

Some records include a system prompt as an additional turn with "from": "system".

Deduplication

Deduplicated using a smart key combining source + source_id + MD5 hash of conversation content. Ensures cross-source CVE coverage is retained while true duplicates are removed.

Intended Use

• Fine-tuning LLMs for cybersecurity reasoning, vulnerability analysis, and threat intelligence
• SOC assistant models
• Security-aware coding assistants
• CTF / penetration testing knowledge bases

Licence

Apache 2.0. Individual source datasets retain their original licences — all sources used are open/public. See source links for individual licence details.

Sources & Attribution

• NVD: https://nvd.nist.gov (public domain)
• OSV: https://osv.dev (various open licences per ecosystem)
• GitHub Advisory Database: https://github.com/github/advisory-database (CC-BY 4.0)
• Security Stack Exchange: https://stackexchange.com/legal (CC-BY-SA 4.0)
• ExploitDB: https://exploit-db.com (public)
• MITRE ATT&CK: https://attack.mitre.org (Apache 2.0)
• CISA KEV: https://cisa.gov/known-exploited-vulnerabilities-catalog (public domain)