CyberSecurity-1M

A large-scale, multi-source cybersecurity knowledge dataset containing 1.1M+ records across 16 categories and 135 sources, covering vulnerability research, threat intelligence, penetration testing, CTF writeups, malware analysis, and more.

Dataset Summary

CyberSecurity-1M aggregates publicly available cybersecurity content from 135 diverse sources into a unified, categorized, and quality-annotated dataset. It is designed to support cybersecurity research, LLM fine-tuning for security domains, threat intelligence analysis, and security education.

The dataset covers the full spectrum of cybersecurity knowledge:

• Vulnerability databases (CVE, exploit collections, advisories)
• Threat intelligence (APT reports, malware analysis, IOCs)
• Offensive security (penetration testing, red teaming, exploit development)
• Defensive security (incident response, detection rules, forensics)
• Security frameworks (MITRE ATT&CK, Sigma rules, Atomic Red Team)
• CTF & training (CTF writeups, exercises, tutorials)
• Security tools (Nuclei templates, Metasploit modules, exploit code)
• Reference materials (cheat sheets, awesome lists, documentation)
• Chinese security community (先知, 安全客, 看雪, FreeBuf, etc.)
• Books & conference talks (OCR-extracted PDFs, DEF CON/Black Hat presentations)

Supported Tasks

• Language modeling / text generation: Pre-train or fine-tune LLMs on cybersecurity domain text
• Summarization: Generate summaries of threat reports, vulnerability advisories
• Question answering: Build cybersecurity QA systems over the knowledge base
• Text classification: Categorize security content by type, severity, or topic
• Information extraction: Extract IOCs, CVEs, TTPs from unstructured text
• Retrieval-augmented generation (RAG): Use as a knowledge base for security-focused chatbots

Dataset Structure

CyberSecurity-1M/
├── merged/                          # Categorized & merged data (16 categories)
│   ├── vulnerability.jsonl          # 876K records — CVE/exploit databases
│   ├── ctf.jsonl                    # 74K records — CTF writeups & exercises
│   ├── cn_sec.jsonl                 # 50K records — Chinese security community
│   ├── reference.jsonl              # 48K records — Cheat sheets, docs, lists
│   ├── tool.jsonl                   # 17K records — Security tools & templates
│   ├── incident_response.jsonl      # 13K records — IR & forensics
│   ├── framework.jsonl              # 7.8K records — MITRE ATT&CK, Sigma, Atomic Red Team
│   ├── bug_bounty.jsonl             # 6.4K records — Bug bounty writeups
│   ├── threat_intel.jsonl           # 5.1K records — APT reports & threat research
│   ├── offsec.jsonl                 # 4.9K records — Offensive security / red team
│   ├── books.jsonl                  # 3.3K records — OCR-extracted cybersecurity books
│   ├── vuln_research.jsonl          # 1.8K records — Vulnerability research blogs
│   ├── news.jsonl                   # 846 records — Security news
│   ├── ics_ot.jsonl                 # 767 records — ICS/OT/SCADA security
│   ├── conference.jsonl             # 353 records — Conference presentations
│   └── ai_security.jsonl            # 207 records — LLM/AI security
└── source_registry.json             # Full inventory of all 135 sources with quality status

Data Instances

Each JSONL record follows the CyberRecord schema:

{
  "id": "a1b2c3d4e5f6",
  "title": "CVE-2024-1234: Remote Code Execution in Framework X",
  "source": "nvd_cve",
  "_original_source": "nvd_cve",
  "url": "https://nvd.nist.gov/vuln/detail/CVE-2024-1234",
  "category": "vulnerability",
  "cve": "CVE-2024-1234",
  "author": null,
  "date": "2024-03-15",
  "tags": ["rce", "critical", "framework-x"],
  "description": "A remote code execution vulnerability exists in...",
  "markdown": "# CVE-2024-1234\n\n## Description\nA remote code execution vulnerability...",
  "exploit_code": null,
  "bounty": null,
  "extra": {},
  "scraped_at": "2026-05-08T12:00:00+00:00",
  "schema_version": 1
}

Data Fields

Field	Type	Description
id	string	Unique record identifier (MD5 hash of URL or composite key)
title	string	Record title
source	string	Original source identifier (e.g., "nvd_cve", "unit42")
_original_source	string	Same as source; preserved for provenance after merging
url	string	Original URL of the content
category	string	Merged category name (one of 16 categories)
cve	string	CVE identifier, if applicable
author	string	Author name(s)
date	string	Publication date
tags	list[string]	Content tags/topics
description	string	Brief description/summary
markdown	string	Full text content in markdown format
exploit_code	string	Source code / payload content (for exploit/tool sources)
bounty	string	Bug bounty amount (for bug_bounty category)
extra	object	Source-specific metadata (varies by source)
scraped_at	string	ISO 8601 timestamp of when the record was scraped
schema_version	int	Schema version (currently 1)

Data Splits

This dataset uses a single train split. Records are organized into 16 category-based configurations (see configs in YAML metadata). Each configuration can be loaded independently:

from datasets import load_dataset

# Load a specific category
ds = load_dataset("morinoppp/CyberSecurity-1M", "vulnerability", split="train")

# Load all categories
ds = load_dataset("morinoppp/CyberSecurity-1M", split="train")

Category Overview

Category	Records	Size	Sources	Description
vulnerability	876K	3.1GB	13	CVE databases, exploit collections, security advisories
ctf	74K	1.9GB	8	CTF writeups, exercises, tutorials
cn_sec	50K	515MB	15	Chinese security community (先知, 安全客, 看雪, etc.)
reference	48K	626MB	13	Cheat sheets, awesome lists, documentation
tool	17K	146MB	13	Security tools, Nuclei templates, Metasploit modules
incident_response	13K	64MB	3	Incident response, forensics, breach databases
framework	7.8K	1.2GB	6	MITRE ATT&CK, Sigma rules, Atomic Red Team
bug_bounty	6.4K	255MB	2	Bug bounty writeups and disclosures
threat_intel	5.1K	641MB	24	APT reports, threat research blogs
offsec	4.9K	71MB	12	Offensive security, red teaming, penetration testing
books	3.3K	407MB	2	OCR-extracted cybersecurity books (PDF/EPUB)
vuln_research	1.8K	80MB	16	Vulnerability research blogs and analysis
news	846	68MB	3	Security news and RSS feeds
ics_ot	767	35MB	2	ICS/OT/SCADA security
conference	353	0.2MB	1	DEF CON, Black Hat, and other conference presentations
ai_security	207	2.3MB	2	LLM/AI security research

Source Quality

The source_registry.json contains quality status for each of the 135 sources:

Status	Count	Description
ok	102	Data quality is good
needs_rescrape	3	Markdown content missing, needs re-crawl
incomplete	14	Only partial data scraped (e.g., 1 RSS page)
low_quality	6	Content too short (summary-only feeds)
structural	10	Structured data where short markdown is expected (exploit code, CVE fields)

Data Collection

Data is collected using cyber-collector, a multi-source Python scraper with plugin architecture supporting:

• RSS/Atom feeds with automatic pagination
• HTML scraping via scrapling (async) and curl_cffi (stealth/WAF bypass)
• JavaScript-rendered pages via Playwright headless browser
• PDF extraction via SiliconFlow PaddleOCR-VL-1.5 cloud OCR and markitdown
• GitHub repository cloning and analysis
• REST API consumption with pagination

Considerations

Social Impact

This dataset aggregates publicly available cybersecurity information. It is intended for defensive security research, education, and LLM training. Some content describes vulnerabilities, exploitation techniques, and offensive tools — this information is already publicly available from the original sources.

Licensing

The dataset compilation is released under Apache 2.0. Individual content items retain their original source licensing. Some sources may have specific terms of service or attribution requirements — users should verify licensing for specific content before redistribution.

Biases

• Language bias: English content dominates (80%), with Chinese as secondary (18%)
• Source bias: Content reflects the perspectives and coverage of the original sources
• Recency bias: Web-scraped content is more complete for recent years
• Category imbalance: Vulnerability data (876K) vastly outnumbers other categories

Citation

@dataset{cybersecurity-1m,
  title={CyberSecurity-1M: A Large-Scale Multi-Source Cybersecurity Knowledge Dataset},
  author={morinoppp},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/morinoppp/CyberSecurity-1M}
}