# HolyCyber — Layer: `detection_engineering` A focused dataset for the `detection_engineering` domain of the HolyCyber corpus, built on its own from that domain's authoritative sources only. ## Overview - **Rows:** 7,970 - **Groups (leakage units):** 4,241 - **Entities in the domain:** 4,286 - **Splits:** test=806, train=6,399, val=765 - **Task types:** 6 | **Subdomains:** 6 - **Row size (tiktoken cl100k, user+assistant):** min 2,085 / median 2,213 / mean 2,254 / p90 2,435 / max 5,006 tokens - **Entity's own material:** median 12% of an answer's tokens; 6,747 rows are below 25% (the balance is the labelled reference appendix) - **Total tokens:** 17,964,031 | **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": "detection_engineering", "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 | 3,694 | | Detection logic | 2,249 | | Platforms | 2,249 | | Engineering note | 2,249 | | Supports technique(s) | 2,232 | | Telemetry | 2,210 | | Telemetry required | 2,177 | | Tunable elements | 2,126 | | Detection strategies and analytics | 1,600 | | Tracked users (groups, malware, tools) | 1,339 | | Analytics | 1,261 | | Detects | 915 | ## 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 | |---|---| | detection_engineering | 4,516 | | control_mapping | 1,294 | | knowledge_explanation | 850 | | query_generation | 724 | | threat_attribution | 534 | | relationship_reasoning | 52 | ## Subdomains | Subdomain | Rows | |---|---| | detection_strategy | 4,318 | | logsource_mapping | 1,956 | | telemetry_coverage | 922 | | rule_explanation | 720 | | rule_tuning | 52 | | query_generation | 2 | ## Sources | Source | Rows | |---|---| | mitre-attack | 7,890 | | arxiv-corpus | 71 | | nist-publication | 8 | | owasp-cheatsheet | 1 | ## Quality controls - **Exact duplicates removed:** 0 - **Near-duplicates removed:** 50 (MinHash-LSH, Jaccard >= 0.90, same task type only) - **Quality gate:** 34 rejected ({'unresolved_markdown_link': 31, 'unbalanced_code_fence': 3}) - **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="detection_engineering/train.jsonl", validation="detection_engineering/val.jsonl", test="detection_engineering/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 - 6,747 of 7,970 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.