README.md

---
license: apache-2.0
task_categories:
  - text-generation
  - text-retrieval
language:
  - en
tags:
  - cybersecurity
  - vulnerability
  - cve
  - threat-intelligence
  - mitre-attack
  - detection-rules
  - ctf
  - incident-response
  - offensive-security
  - books
  - public-domain
  - jsonl
size_categories:
  - 1M<n<10M
pretty_name: CyberSecurity-Corpus
---

# CyberSecurity-Corpus

Public **CyberRecord** JSONL corpus of open, licence-safe cybersecurity knowledge for research and training.

**This dataset was built from scratch from public feeds and open-source repositories.**  
It does **not** copy or derive from gated datasets such as `WhitzardAgent/CyberSecurity-1M`.

## Schema (CyberRecord, `schema_version: 1`)

Each JSONL line is one record:

| Field | Type | Notes |
|-------|------|-------|
| `id` | string | 12-char stable hash of `source\|url\|cve-or-key` |
| `title` | string | |
| `source` | string | short source id |
| `_original_source` | string | same as `source` |
| `url` | string | canonical URL |
| `category` | string | see below |
| `cve` | string\|null | `CVE-YYYY-NNNN` when known |
| `author` | string\|null | |
| `date` | string\|null | `YYYY-MM-DD` |
| `tags` | string[] | |
| `description` | string | brief |
| `markdown` | string | full markdown body |
| `exploit_code` | string\|null | e.g. nuclei template body when applicable |
| `bounty` | null | unused |
| `extra` | object | source-specific metadata |
| `scraped_at` | string | ISO8601 UTC |
| `schema_version` | int | `1` |

### Categories

`vulnerability`, `framework`, `reference`, `tool`, `threat_intel`, `vuln_research`, `incident_response`, `offsec`, `ctf`, `bug_bounty`, `conference`, `news`, `ai_security`, `ics_ot`, `books`

(Only categories with data appear in `data/merged/`.)

## Layout

```
data/merged/{category}.jsonl   # primary splits by category
data/by_source/{source}.jsonl  # same records grouped by source
sources/LICENCES.md            # per-source licence notes
sources/books/                 # downloaded PD / USG book PDFs & texts
scripts/                       # reproducible build scripts
```

## Record counts (this build)

**Total: 2,328,955**

### By category

| Category | Count |
|----------|------:|
| `vulnerability` | 2,295,621 |
| `threat_intel` | 13,685 |
| `tool` | 7,738 |
| `framework` | 6,056 |
| `incident_response` | 2,110 |
| `vuln_research` | 1,724 |
| `ctf` | 903 |
| `offsec` | 473 |
| `books` | 350 |
| `ics_ot` | 171 |
| `reference` | 124 |

### By source

| Source | Count |
|--------|------:|
| `osv` | 1,892,190 |
| `cvelistV5` | 398,867 |
| `nuclei_templates` | 14,026 |
| `threatfox` | 11,957 |
| `sigma` | 3,144 |
| `mitre_attack` | 2,257 |
| `cisa_kev` | 1,728 |
| `azure_sentinel` | 1,388 |
| `capec` | 615 |
| `atomic_red_team` | 344 |
| `ctf_0e85` | 280 |
| `ad_exploit_cheatsheet` | 245 |
| `owasp_asvs` | 211 |
| `atc` | 210 |
| `ctflearn_writeups` | 173 |
| `payloadsallthethings` | 141 |
| `owasp_cheatsheets` | 124 |
| `csivitu_ctf` | 109 |
| `washi_ctf` | 104 |
| `nist_sp800_53r5_book` | 94 |
| `nist_sp800_61` | 80 |
| `cloudgoat` | 66 |
| `htb_writeups` | 65 |
| `gutenberg_75074` | 57 |
| `pdkt_ctf` | 51 |
| `sajjadium_ctf` | 50 |
| `cisa_ir_playbooks` | 44 |
| `pagerduty_ir` | 36 |
| `cfi_ul_ctf` | 34 |
| `nist_sp800_12_book` | 33 |
| `friedman_military_cryptanalysis_p4` | 26 |
| `nist_sp800_100_book` | 23 |
| `friedman_elements_cryptanalysis` | 21 |
| `ceh_notes` | 20 |
| `gutenberg_48232` | 18 |
| `friedman_military_cryptanalysis_p1` | 18 |
| `kernelpwn` | 17 |
| `friedman_military_cryptanalysis_p2` | 17 |
| `nist_sp800_12r1_book` | 15 |
| `gutenberg_48871` | 14 |
| `jibeee_ctf` | 10 |
| `gutenberg_43515` | 10 |
| `palantir_ads` | 8 |
| `oriapp_writeups` | 6 |
| `pasticciotto` | 4 |
| `gutenberg_26839_ch11` | 4 |
| `secret_knowledge` | 1 |

Built at `2026-09-29T01:47:44Z` (UTC).

### Recently expanded: `books` (public-domain / USG / CC0)

Pure modern cybersecurity textbooks that are clearly PD/CC0 are scarce. This release therefore ships:

1. Classic **public-domain cryptanalysis / cipher** manuals (Project Gutenberg + NSA/Friedman USG CC0 OCR)
2. Closely related **military signaling / telegraph code** PD works
3. Longer **NIST handbook-style** Special Publications (US Government works), chunked as books (~4096 `cl100k_base` tokens)

| Book | Author | Licence | Chunks |
|------|--------|---------|-------:|
| Elementary Cryptanalysis | Helen Fouché Gaines | Public domain (USA) | 57 |
| Manual for the Solution of Military Ciphers | Parker Hitt | Public domain (USA) | 14 |
| The Robinson Telegraphic Cipher | Stephen L. Robinson | Public domain (USA) | 18 |
| Visual Signaling | United States Army Signal Corps | Public domain (USA) | 10 |
| Mathematical Recreations and Essays — Chapter XI: Cryptographs and Ciphers | W. W. Rouse Ball | Public domain (USA) | 4 |
| Elements of Cryptanalysis (Training Pamphlet No. 3) | William F. Friedman | CC0 1.0 (US Government work; Internet Archive) | 21 |
| Military Cryptanalysis, Part I: Monoalphabetic Substitution Systems (3rd ed.) | William F. Friedman | CC0 1.0 (US Government work; Internet Archive) | 18 |
| Military Cryptanalysis, Part II: Simpler Varieties of Polyalphabetic Substitution Systems | William F. Friedman | CC0 1.0 (US Government work; Internet Archive) | 17 |
| Military Cryptanalysis, Part IV: Transposition and Fractionating Systems | William F. Friedman | CC0 1.0 (US Government work; Internet Archive) | 26 |
| NIST SP 800-12 Rev. 1: An Introduction to Information Security | NIST (Nieles, Dempsey, Pillitteri) | US Government work (not subject to copyright in the US) | 15 |
| NIST SP 800-12: An Introduction to Computer Security — The NIST Handbook | NIST (Guttman, Roback) | US Government work (not subject to copyright in the US) | 33 |
| NIST SP 800-100: Information Security Handbook — A Guide for Managers | NIST | US Government work (not subject to copyright in the US) | 23 |
| NIST SP 800-53 Rev. 5: Security and Privacy Controls for Information Systems and Organizations | NIST | US Government work (not subject to copyright in the US) | 94 |

**New `books` records:** 350  
Tokenizer: `cl100k_base` target ~4096 tokens/chunk.  
Friedman volumes use Internet Archive DjVu OCR — expect scan artefacts (tagged `ocr`).

### Previously expanded categories

| Category | Count | Notes |
|----------|------:|-------|
| `incident_response` | 2,110 | CISA/NIST playbooks, PagerDuty IR docs, Atomic Red Team, ATC response, Azure Sentinel hunting/detections |
| `ctf` | 903 | MIT/Apache CTF writeup & challenge repos from GitHub |
| `offsec` | 473 | PayloadAllTheThings, AD exploitation cheat sheet, CloudGoat labs, CEH notes (educational; no malware packaging) |

## Sources & licences

Compilation packaging: **Apache-2.0**. Individual records remain under their upstream terms — see `sources/LICENCES.md`.

**Excluded on purpose:** gated/NC datasets (incl. HackTricks CC BY-NC), GPL copyleft knowledge bases (GTFOBins/LOLBAS/ScoutSuite), Elastic License rules, vendor blog scrapes without redistribution rights, HackerOne bulk scrapes, cracked course dumps, copyrighted modern security textbooks (O’Reilly etc.), All-Rights-Reserved “free PDF” copies.

## Intended use

- Retrieval / RAG corpora for defensive security assistants
- Classification and tagging experiments
- CVE / ATT&CK / detection-rule linking research
- CTF writeup / IR playbook / educational offsec note retrieval
- Historical cryptanalysis / NIST handbook retrieval (`books`)

**Not** a substitute for official CVE/NVD/OSV APIs for production vulnerability management. Always verify against primary sources. Offsec records are educational reference text, not attack tooling.

## Rebuild

```bash
# See scripts/build_records.py, scripts/append_osv_threatfox.py,
# scripts/ingest_ctf_ir_offsec.py, scripts/ingest_offsec_extra.py,
# scripts/ingest_books_pd.py
# Requires clones/downloads under tmp/ and sources/books/ as documented in sources/LICENCES.md
```

## Citation

If you use this corpus, please cite the upstream projects above and this dataset repo.