# HolyCyber — Layer: `cloud_security`

A focused dataset for the `cloud_security` domain of the HolyCyber corpus, built on its own from that domain's authoritative sources only.

## Overview

- **Rows:** 2,254
- **Groups (leakage units):** 590
- **Entities in the domain:** 590
- **Splits:** test=218, train=1,804, val=232
- **Task types:** 7  |  **Subdomains:** 5
- **Row size (tiktoken cl100k, user+assistant):** min 2,076 / median 2,313 / mean 2,356 / p90 2,521 / max 5,288 tokens
- **Entity's own material:** median 18% of an answer's tokens; 1,548 rows are below 25% (the balance is the labelled reference appendix)
- **Total tokens:** 5,310,326  |  **Rows below the 2,000-token floor:** 0
- **Canary:** every row carries the decontamination canary

## Record format

One JSON object per line; `messages` is a three-turn chat (system / user / assistant).

```json
{"id": "6d408e38f3cb8dff", "row_key": "<domain>:<group_key>:<task_type>:<subdomain>:<part>",
 "domain": "cloud_security", "subdomain": "...",
 "task_type": "...", "difficulty": "intermediate", "group_key": "...",
 "source": "...", "license": "...", "split": "train", "canary": "...",
 "tokens": {"user": 180, "assistant": 2350, "total": 2530},
 "core_share": 0.41, "appendix": true,
 "references": [{"title": "...", "url": "..."}],
 "messages": [{"role": "system", "content": "..."},
              {"role": "user", "content": "..."},
              {"role": "assistant", "content": "## ...\n..."}]}
```

| Field | Meaning |
|---|---|
| `id` | content hash of user+assistant, stable across rebuilds of the same row |
| `row_key` | stable identity across rebuilds (`domain:group:task:subdomain:part`) |
| `domain` / `subdomain` | this layer, and the slice of it the row covers |
| `task_type` | the question shape (detection, mitigation, mapping, ...) |
| `difficulty` | derived from the task shape, not guessed |
| `group_key` | the source entity; the unit of the train/val/test split |
| `source` / `license` | the authoritative source, and its licence |
| `tokens` | measured `cl100k_base` counts for user, assistant and total |
| `core_share` | share of the answer that is the entity's own material (1 - appendix) |
| `appendix` | whether the row carries a `## Reference:` appendix |
| `references` | resolvable links for the entities cited in the answer |
| `canary` | decontamination marker |

## What a row contains

Topic sections used in this layer's answers (structural labels such as `Answer` or `Context` and the appendix are excluded):

| Section | Rows |
|---|---|
| Summary | 2,106 |
| Detection strategies and analytics | 945 |
| Tracked users (groups, malware, tools) | 818 |
| Mitigations | 720 |
| Background | 557 |
| Classification | 543 |
| Scope note | 543 |
| Relevant mitigations | 458 |
| Sub-techniques | 85 |
| Introduction | 26 |
| Abstract | 15 |
| Conclusion | 11 |

## How this layer is built

1. **Base cells** — the rows the source parsers emit for this domain (one per entity x task type x subdomain).
2. **Task shapes** — the same entity asked a different defensive question (detection, mitigation, framework mapping, response), built only from other sections of that entity's own record. No new facts are introduced.
3. **Reference pack** — a row still under the token floor is padded with compact descriptions of other entities *of this domain*, taken from this domain's own source text; the window rotates per row.

Every row therefore comes from this domain's sources only, clears the 2,000-token floor, and no two rows are the same.


## Task types

| Task type | Rows |
|---|---|
| knowledge_explanation | 693 |
| control_mapping | 543 |
| detection_engineering | 405 |
| threat_attribution | 346 |
| mitigation_remediation | 261 |
| definition | 5 |
| ai_security_analysis | 1 |

## Subdomains

| Subdomain | Rows |
|---|---|
| audit_logging | 751 |
| compute_abuse | 696 |
| container_security | 544 |
| identity_posture | 262 |
| saas_abuse | 1 |

## Sources

| Source | Rows |
|---|---|
| mitre-attack | 2,103 |
| arxiv-corpus | 118 |
| ietf-rfc | 14 |
| nist-publication | 11 |
| owasp-cheatsheet | 5 |
| research-corpus | 3 |

## Quality controls

- **Exact duplicates removed:** 0
- **Near-duplicates removed:** 1 (MinHash-LSH, Jaccard >= 0.90, same task type only)
- **Quality gate:** 2 rejected ({'unresolved_markdown_link': 2})
- **Contaminated rows removed:** 0 ({})
- **Split integrity:** group-level by `group_key`; no group crosses a split

## Intended use

Supervised fine-tuning / evaluation of a defensive cybersecurity assistant specialised in this domain. All rows are defensive, detection or analysis oriented; none contains functional exploit code or weaponised prompts.

```python
from datasets import load_dataset
ds = load_dataset("json", data_files=dict(train="cloud_security/train.jsonl", validation="cloud_security/val.jsonl", test="cloud_security/test.jsonl"))
# rows whose answer is mostly the entity itself, not the appendix
core = ds["train"].filter(lambda r: r["core_share"] >= 0.5)
```

## Limitations

- 1,548 of 2,254 rows are more than 75% labelled reference appendix: this domain's own records are shorter than the 2,000-token floor, so the row is completed with other entities' records. Mask the appendix when fine-tuning if you only want the core answer.

- Paragraph-level reuse is measured in `research/AUDIT_REPORT.md`; rows are unique but their appendix material is shared.