This dataset is available exclusively for academic and non-commercial research. Please provide your institutional email and affiliation below. Requests from non-institutional email addresses will not be approved. Review typically takes 1-2 business days.

This dataset is provided strictly for academic, non-commercial research purposes only. It contains cybersecurity content including vulnerability details, exploit techniques, and offensive security information. By requesting access, you confirm that you will use this dataset solely for lawful academic research, education, or defensive security purposes, and will not use it for any commercial or illegal activities. Access requests using personal email addresses (Gmail, Outlook, QQ, etc.) will be rejected — please use your institutional/organizational email.

CyberSecurity-1M

A large-scale, multi-source cybersecurity knowledge dataset containing 1.19M+ records across 16 categories, collected exclusively for academic, non-commercial research purposes.

Disclaimer: This dataset is provided for academic research only. All content is aggregated from publicly available sources. The views, opinions, and information expressed in the dataset content do not represent the views or positions of the research team. The research team does not endorse, support, or take responsibility for any of the content. Users access and use this dataset at their own risk.

Security Notice: This dataset contains information about cybersecurity vulnerabilities, exploitation techniques, and offensive security methods. This information is already publicly available and is collected here solely for defensive security research and education. Misuse of this information to attack systems without authorization is illegal. Users must comply with all applicable laws and regulations.

Dataset Summary

CyberSecurity-1M aggregates publicly available cybersecurity content from 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 records, security advisories)
Threat intelligence (APT reports, malware analysis, IOCs)
Offensive security (penetration testing, red teaming)
Defensive security (incident response, detection rules, forensics)
Security frameworks (attack frameworks, detection rulesets)
CTF & training (CTF writeups, exercises, tutorials)
Security tools (templates, modules, exploit code)
Reference materials (cheat sheets, documentation, curated lists)
Chinese-language security community content
Books & conference talks (OCR-extracted PDFs, presentation transcripts)

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 systems

Dataset Structure

CyberSecurity-1M/
├── merged/                          # Categorized & merged data (16 categories)
│   ├── vulnerability.jsonl          # ~879K records
│   ├── cn_sec.jsonl                 # ~57K records
│   ├── reference.jsonl              # ~44K records
│   ├── framework.jsonl              # ~56K records
│   ├── ctf.jsonl                    # ~43K records
│   ├── tool.jsonl                   # ~18K records
│   ├── incident_response.jsonl      # ~18K records
│   ├── bug_bounty.jsonl             # ~17K records
│   ├── offsec.jsonl                 # ~10K records
│   ├── threat_intel.jsonl           # ~13K records
│   ├── conference.jsonl             # ~4.5K records
│   ├── news.jsonl                   # ~2.1K records
│   ├── vuln_research.jsonl          # ~19K records
│   ├── books.jsonl                  # ~3.1K records
│   ├── ai_security.jsonl            # ~1.8K records
│   └── ics_ot.jsonl                 # ~944 records
└── source_registry.json             # Inventory of all 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": "vuln_db",
  "_original_source": "vuln_db",
  "url": "https://example.com/advisory/CVE-2024-1234",
  "category": "vulnerability",
  "cve": "CVE-2024-1234",
  "author": null,
  "date": "2024-03-15",
  "tags": ["rce", "critical"],
  "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
`title`	string	Record title
`source`	string	Source identifier
`_original_source`	string	Same as source; preserved for provenance after merging
`url`	string	Original URL of the content
`category`	string	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 tool sources)
`bounty`	string	Bug bounty amount (for bug_bounty category)
`extra`	object	Source-specific metadata
`scraped_at`	string	ISO 8601 timestamp of when the record was collected
`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("WhitzardAgent/CyberSecurity-1M", "vulnerability", split="train")

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

Category Overview

Category	Records	Size	Description
vulnerability	~879K	3.4GB	CVE records, security advisories, exploit databases
cn_sec	~57K	653MB	Chinese-language security content
framework	~58K	750MB	Security frameworks, detection rules, standards
reference	~44K	984MB	Documentation, cheat sheets, curated lists
ctf	~43K	1.7GB	CTF writeups and exercises
tool	~18K	168MB	Security tools and templates
incident_response	~18K	95MB	Incident response, DFIR, forensics, security log data
bug_bounty	~17K	281MB	Bug bounty writeups and HackerOne disclosures
offsec	~10K	182MB	Offensive security and penetration testing
threat_intel	~13K	1.1GB	Threat research, APT reports, IOCs (ThreatFox, MalwareBazaar, URLhaus, MISP)
vuln_research	~19K	520MB	Vulnerability research, CWE, detection rules (YARA, Elastic, Sigma)
conference	~4.5K	352MB	Security conference papers and presentations
books	~3.1K	315MB	OCR-extracted cybersecurity books
news	~2.1K	176MB	Security news and podcasts
ai_security	~1.8K	14MB	LLM/AI security, Web3 security
ics_ot	~944	38MB	ICS/OT/SCADA security

Category Content Types

Each category contains different types of knowledge. Understanding these types helps select the right data for your use case:

Category	Knowledge Type	Content Characteristics	Example Sources
vulnerability	Structured factual data	CVE records with standardized fields (CVSS scores, affected products, references). Highly structured, ID-keyed, minimal prose. Good for lookup, classification, and vulnerability assessment training.	NVD, OSV, ExploitDB, GitHub Advisories, CISA KEV, vendor security bulletins
cn_sec	Mixed (articles + tutorials)	Chinese-language security articles spanning vulnerability analysis, tool tutorials, and threat commentary. Mix of knowledge-type reference and step-by-step guides. Language is zh-CN.	Anquanke, SecWiki, Xianzhi, Seebug, Govuln, Kanxue, Tttang
reference	Knowledge-type reference	Curated link collections (awesome-lists), cheat sheets, quick-reference cards, and documentation indexes. High link density, concise descriptions, broad coverage. Best for building knowledge graphs and resource discovery.	awesome-security, awesome-web-security, Gtfobins, Lolbas, Seclists, GitHub security markdown
framework	Rule-based structured data	Detection rules (Sigma, YARA, Suricata/Snort ET Open), MITRE ATT&CK mappings, OWASP standards (ASVS, WSTG, CheatSheetSeries, Top10), and security standards/benchmarks (CIS, NIST). Highly structured, machine-parseable format. Ideal for rule generation and detection engineering.	SigmaHQ, MITRE ATT&CK, CAPEC, YARA rules, ET Open rules, capa rules, CIS benchmarks, OWASP projects
ctf	Tutorial/procedural	Step-by-step CTF writeups with detailed exploitation chains, debugging notes, and solution walkthroughs. Rich procedural reasoning — shows how to think about a problem, not just the answer.	CTFtime, CTFSearch, CTF Wiki, LiveOverflow, PentesterLand, Oxdf
tool	Code + documentation	Security tool configurations (Nuclei templates, Metasploit modules), scanner rules, and tool documentation. Mixed code-and-prose format. Useful for tool usage training and rule authoring.	Nuclei templates, Metasploit, Loldrivers, CloudGoat, Kubesploit
incident_response	Procedural + knowledge + log data	DFIR playbooks, forensic analysis guides, memory forensics documentation, attack sample datasets (EVTX, PCAP), and security log data (Splunk BOTS, Mordor, EVTX-ATTACK-SAMPLES). Mix of step-by-step procedures, reference knowledge, and raw log samples for training.	Volatility, FLARE-FLOSS, Mordor datasets, Malware Traffic Analysis, DFIR KB, Splunk BOTS, Security Log Data
offsec	Tutorial/procedural	Red team infrastructure guides, penetration testing methodology docs, attack chain walkthroughs, and offensive tool usage. Strong procedural content showing step-by-step exploitation.	HackTricks, IRedTeam, PentestBook, PSTips, Offensive Security KB, Red Team Cheatsheets
threat_intel	Analytical/report + IOC data	APT campaign analysis, malware family reports, threat landscape assessments, and structured IOC collections (ThreatFox, MalwareBazaar, URLhaus, MISP OSINT feeds). Mix of long-form analytical prose with TTP mapping and structured indicator data. Best for threat reasoning, report summarization, and IOC extraction.	Unit42, Kaspersky, CrowdStrike, ESET, Check Point, Mandiant, ThreatFox, MalwareBazaar, URLhaus, MISP feeds
bug_bounty	Tutorial/procedural	Bug bounty writeups with reproduction steps, bypass techniques, and bounty amounts. Step-by-step vulnerability discovery narratives. Excellent for vulnerability discovery training.	PentesterLand, HackerOne Hacktivity, Bug Bounty KB
vuln_research	Analytical/deep-dive + rule data	In-depth vulnerability analysis, exploit development writeups, reverse engineering case studies, CWE weakness taxonomy, and detection rules (YARA, Elastic, Sigma, capa). Mix of analytical prose and structured rule/code content. Shows why a vulnerability exists and how it was found and detected.	Trail of Bits, PortSwigger Research, Google Project Zero, GH Security Lab, full-disclosure, CWE KB, YARA Rules, Elastic Detection Rules, Sigma
books	Long-form knowledge	Full-length cybersecurity books (OCR-extracted from PDFs). Covers topics from network security to cryptography to forensics. Long coherent text ideal for domain pre-training.	Local book collection
conference	Academic/presentation	Security conference papers (USENIX Security, NDSS) and presentation slides (DEF CON, BlackHat). Mix of rigorous academic prose and slide-format bullet points.	USENIX papers, NDSS papers, DEF CON media, security conference repos
news	Factual/news	Security industry news articles, breach reports, and podcast transcripts. Timely, event-driven content. Good for security awareness and event classification.	Krebs on Security, Security Now, Dark Reading, Darknet Diaries, BleepingComputer
ai_security	Research/tutorial	LLM security research (jailbreaks, adversarial prompts), AI vulnerability scanning tools, and Web3/smart-contract security. Emerging domain with mixed analytical and procedural content.	LLM-attacks, NVIDIA garak, GPTFuzz, Guardrails, Web3 Security KB
ics_ot	Specialized reference	Industrial control system security: SCADA protocols, OT pentesting guides, ICS-specific detection rules. Niche domain with specialized terminology and procedures.	ICS Security Tools, Redpoint, ThreatHunting Keywords, GRASSMARLIN

Content Type Legend

Structured factual data: Machine-parseable records with standardized fields (CVE, CVSS, CPE). Low prose density, high lookup value.
Knowledge-type reference: Curated lists, documentation, cheat sheets. Broad coverage, concise entries, high link density.
Rule-based structured data: Detection rules, security policies, compliance benchmarks. Machine-parseable format (YAML, Sigma, YARA).
Tutorial/procedural: Step-by-step walkthroughs, exploitation chains, and debugging narratives. Shows how to accomplish a task.
Analytical/report: Threat research reports, APT analysis, and vulnerability deep-dives. Long-form reasoning about why and what it means.
Academic/presentation: Peer-reviewed papers and conference slides. Rigorous methodology, formal language.
Long-form knowledge: Full books and extended reference works. Coherent narrative across chapters.
Factual/news: Time-bound event reporting, breach disclosures, industry updates.
Specialized reference: Domain-specific (ICS/OT) reference materials with niche terminology.
Research/tutorial: Emerging domain content mixing analytical findings and practical techniques.

Considerations

Academic Use Only

This dataset is compiled and distributed strictly for academic, non-commercial research purposes. Any commercial use, redistribution for profit, or application in commercial products is strictly prohibited without explicit written authorization. The research team receives no financial benefit from this dataset.

Disclaimer

The content in this dataset is aggregated from publicly available sources and represents the views of the original authors, not the research team. The research team:

Does not endorse, verify, or guarantee the accuracy of any content
Does not take responsibility for any claims, opinions, or information in the dataset
Does not encourage or support the use of this information for unauthorized access or illegal activities
Makes no warranties, express or implied, regarding the dataset's fitness for any particular purpose

Security Risk Notice

This dataset contains technical information about vulnerabilities, exploitation methods, and offensive security techniques. While this information is already publicly available, users should be aware that:

Unauthorized use of exploit techniques against systems you do not own or have explicit permission to test is illegal in most jurisdictions
Responsible disclosure practices should be followed when discovering new vulnerabilities
Users must comply with all applicable local, national, and international laws
The dataset should only be used to improve defensive security capabilities

Licensing

The dataset compilation is released under Apache 2.0 for academic, non-commercial use. Individual content items retain their original source licensing. Some content may have specific terms of service or attribution requirements — users must verify licensing for specific content before any redistribution. Commercial use is prohibited.

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: Content is more complete for recent years
Category imbalance: Vulnerability data (~~879K) dominates at ~74%, though threat_intel (~~13K), vuln_research (~~19K), and framework (~~58K) have grown significantly with new IOC data, detection rules, and CWE/OWASP content

Citation

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

Downloads last month: 341

Curated by:

WhitzardAgent

License:

academic, non-commercial use only

Languages:

English (primary), Chinese (secondary)

Size:

~10.5GB, 1.19M+ records