README.md

---
task_categories:
- text-classification
language:
- en
license: cc-by-4.0
library_name: datasets
tags:
- vulnerability
- cybersecurity
- security
- cve
- mitre-attack
- attack-techniques
dataset_info:
  features:
  - name: id
    dtype: string
  - name: title
    dtype: string
  - name: description
    dtype: string
  - name: exploitation_techniques
    list: string
  - name: primary_impact
    list: string
  - name: secondary_impact
    list: string
  - name: techniques
    list: string
  - name: techniques_derived
    list: string
  - name: label_sources
    list: string
  - name: attack_version
    dtype: string
  - name: cvss_vector
    dtype: string
  - name: cvss_version
    dtype: string
  - name: cwes
    list: string
  - name: affected_products
    list: string
  - name: cpes
    list: string
  splits:
  - name: train
    num_bytes: 2228443
    num_examples: 1086
  - name: test
    num_bytes: 248288
    num_examples: 121
  download_size: 2142998
  dataset_size: 2476731
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
---

# vulnerability-attack-techniques

This dataset maps **1,207 CVEs** to **MITRE ATT&CK (Enterprise) techniques**, joining
hand-curated mappings from the [MITRE Center for Threat-Informed Defense (CTID)](https://ctid.mitre.org/)
with vulnerability descriptions from
[CIRCL/vulnerability-scores](https://huggingface.co/datasets/CIRCL/vulnerability-scores).
It is intended for training and evaluating models that suggest candidate ATT&CK
techniques from a vulnerability description: CVSS tells you *how bad* a
vulnerability is, CWE *what kind of flaw* it is — ATT&CK tells defenders *what
adversary behavior to expect and detect*.

Every label in the `techniques` column was written by an analyst following the CTID
["Mapping ATT&CK to CVE for Impact" methodology](https://github.com/center-for-threat-informed-defense/attack_to_cve/blob/master/methodology.md),
which assigns each CVE up to three kinds of techniques: an **exploitation
technique** (how it is exploited), a **primary impact** (what exploitation
directly yields), and a **secondary impact** (what the attacker can do next).

This is the gold set of the paper
[*Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and
the Limits of LLM-Assisted Label Expansion*](https://arxiv.org/abs/2607.25572)
(arXiv:2607.25572). The classifier trained on it,
[CIRCL/vulnerability-attack-technique-classification-roberta-base](https://huggingface.co/CIRCL/vulnerability-attack-technique-classification-roberta-base),
runs in production on [Vulnerability-Lookup](https://vulnerability.circl.lu).

DOI: [10.57967/hf/9621](https://doi.org/10.57967/hf/9621)

## Label sources

| `label_sources` | CVEs | Origin |
|-----------------|------|--------|
| `ctid_cve` | 788 | [attack_to_cve](https://github.com/center-for-threat-informed-defense/attack_to_cve) (2021), ATT&CK v9 era |
| `ctid_kev` | 392 | [Mappings Explorer](https://center-for-threat-informed-defense.github.io/mappings-explorer/) KEV mappings, ATT&CK 16.1 |
| both | 27 | |

All technique IDs are normalized to **enterprise ATT&CK v19.1**: techniques revoked
since the original mappings are remapped to their successor via the STIX
`revoked-by` relationships (e.g. T1562 *Impair Defenses* → T1685 *Disable or
Modify Tools*), and Mobile/ICS techniques are dropped (enterprise domain only).

## ⚠️ `techniques` vs `techniques_derived`

The `techniques_derived` column contains labels from the automatically derived
CVE → CWE → CAPEC → ATT&CK chain maintained by
[CVE2CAPEC](https://github.com/Galeax/CVE2CAPEC). **Do not train on this
column.** Analysis of the chain shows a median fan-out of 4–20 techniques per
CVE and top-frequency techniques (e.g. T1574.007 on 53% of 2024 CVEs) that are
artifacts of the cross-framework table expansion, not descriptions of real
adversary behavior. The column is included as:

1. a baseline that a trained model must beat;
2. a comparison column for studying where the deterministic chain diverges from analyst judgment.

Its use as an inference-time candidate prior was measured and **rejected**
(2026-08-06): at the parent-technique level the derived candidate sets cover
only 3.3% of the analyst-chosen techniques on the test split, so any
re-ranking toward them degrades every ranking metric.

The full source analysis is documented in the
[VulnTrain documentation](https://github.com/vulnerability-lookup/VulnTrain/blob/main/docs/attack-techniques-dataset.md).

## Fields

| Field | Type | Description |
|-------|------|-------------|
| `id` | string | CVE identifier |
| `title` | string | Vulnerability title |
| `description` | string | Vulnerability description in English (model input) |
| `exploitation_techniques` | list[string] | CTID exploitation technique(s) |
| `primary_impact` | list[string] | CTID primary impact technique(s) |
| `secondary_impact` | list[string] | CTID secondary impact technique(s) |
| `techniques` | list[string] | Union of all curated techniques — the training target |
| `techniques_derived` | list[string] | CVE2CAPEC weak labels — **not** for training |
| `label_sources` | list[string] | `ctid_cve` and/or `ctid_kev` |
| `attack_version` | string | Enterprise ATT&CK version the IDs are normalized to |
| `cvss_vector` | string | CVSS vector string, highest available version (empty if none) — v2 |
| `cvss_version` | string | Version of `cvss_vector`: `4.0`, `3.1`, `3.0` or `2.0` — v2 |
| `cwes` | list[string] | CWE assignments, e.g. `CWE-502 Deserialization of Untrusted Data` — v2 |
| `affected_products` | list[string] | `vendor product` pairs from the CVE record — v2 |
| `cpes` | list[string] | CPE identifiers — v2 |

### Structured metadata columns (v2, added 2026-08-06)

The v2 columns are extracted from the raw CVE records served by
[Vulnerability-Lookup](https://vulnerability.circl.lu) (CNA container
preferred, [CISA ADP Vulnrichment](https://github.com/cisagov/vulnrichment)
filling many gaps — notably 100% CVSS/CWE coverage on the KEV subset);
`cpes` is joined from
[CIRCL/vulnerability-scores](https://huggingface.co/datasets/CIRCL/vulnerability-scores).
v1 columns are unchanged (the update is strictly additive: identical rows
and splits). Coverage differs by label source — report results stratified
by `label_sources` when using these columns as model inputs:

| Subset | CVEs | `cvss_vector` | `cwes` | `affected_products` | `cpes` |
|--------|------|---------------|--------|---------------------|--------|
| all | 1,207 | 72.0% | 84.3% | 67.4% | 93.2% |
| `ctid_kev` | 392 | 100% | 100% | 79.8% | 79.1% |
| `ctid_cve` | 788 | 57.1% | 76.0% | 62.2% | 100% |
| both | 27 | 100% | 100% | 40.7% | 100% |

CVSS versions among the 869 vectors: 677 × v3.1, 173 × v3.0, 18 × v4.0, 1 × v2.0.

## Label statistics

192 distinct techniques; 66 with at least 5 examples. Most CVEs carry 1–3
techniques. Top techniques: T1190 *Exploit Public-Facing Application* (348),
T1059 *Command and Scripting Interpreter* (262), T1203 *Exploitation for Client
Execution* (213), T1068 *Exploitation for Privilege Escalation* (189).

## Known limitations

- **Size**: ~1,200 CVEs supports a proof-of-concept, not a production model.
- **Selection bias**: both label sources over-represent exploited-in-the-wild
  vulnerabilities (the KEV set by construction).
- **Inherent task ceiling**: a CVE description describes a flaw, while ATT&CK
  describes attacker behavior around it — even human annotators disagree on
  such mappings. Models trained on this data should *suggest candidate
  techniques for analyst review*, not produce authoritative mappings.

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("CIRCL/vulnerability-attack-techniques")

for entry in dataset["train"].select(range(3)):
    print(entry["id"], entry["techniques"], "-", entry["description"][:80])
```

## Licensing of upstream sources

The CTID mappings are Apache-2.0. Descriptions come from
[CIRCL/vulnerability-scores](https://huggingface.co/datasets/CIRCL/vulnerability-scores)
(CC BY 4.0). The `techniques_derived` column is derived from the GPLv3
[CVE2CAPEC](https://github.com/Galeax/CVE2CAPEC) project. MITRE ATT&CK® is a
registered trademark of The MITRE Corporation; ATT&CK content is used in
accordance with the [MITRE ATT&CK terms of use](https://attack.mitre.org/resources/legal-and-branding/terms-of-use/).

## Related artifacts

| Artifact | Location | DOI |
|----------|----------|-----|
| Released model trained on this dataset | [CIRCL/vulnerability-attack-technique-classification-roberta-base](https://huggingface.co/CIRCL/vulnerability-attack-technique-classification-roberta-base) | [10.57967/hf/9623](https://doi.org/10.57967/hf/9623) |
| LLM expansion dataset (negative result) | [CIRCL/vulnerability-attack-techniques-llm-scaling](https://huggingface.co/datasets/CIRCL/vulnerability-attack-techniques-llm-scaling) | [10.57967/hf/9622](https://doi.org/10.57967/hf/9622) |
| LLM-expanded comparison model | [CIRCL/vulnerability-attack-technique-classification-roberta-base-llm-expanded](https://huggingface.co/CIRCL/vulnerability-attack-technique-classification-roberta-base-llm-expanded) | [10.57967/hf/9624](https://doi.org/10.57967/hf/9624) |
| Code | [vulnerability-lookup/VulnTrain](https://github.com/vulnerability-lookup/VulnTrain) | — |
| Paper | [arXiv:2607.25572](https://arxiv.org/abs/2607.25572) | — |
| Paper LaTeX source + trainer logs | [vulnerability-lookup/cve-attack-mapping-paper](https://github.com/vulnerability-lookup/cve-attack-mapping-paper) | — |

## References

- [Vulnerability-Lookup](https://vulnerability.circl.lu) — the vulnerability data source
- [VulnTrain](https://github.com/vulnerability-lookup/VulnTrain) — generation pipeline (`vulntrain-dataset-attack-generation`)
- [Methodology documentation](https://github.com/vulnerability-lookup/VulnTrain/blob/main/docs/attack-techniques-dataset.md)
- [MITRE CTID attack_to_cve](https://github.com/center-for-threat-informed-defense/attack_to_cve) and [Mappings Explorer](https://center-for-threat-informed-defense.github.io/mappings-explorer/)
- [CVE2CAPEC](https://github.com/Galeax/CVE2CAPEC) by Galeax

## Citation

```bibtex
@misc{bonhomme2026mappingcvesmitreattck,
      title={Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion},
      author={Cédric Bonhomme and Alexandre Dulaunoy},
      year={2026},
      eprint={2607.25572},
      archivePrefix={arXiv},
      primaryClass={cs.CR},
      url={https://arxiv.org/abs/2607.25572},
}
```

## Acknowledgements

Developed at [CIRCL](https://www.circl.lu) in the context of the
[AIPITCH](https://www.science.nask.pl/en/research-areas/projects/12456)
project, co-funded by the European Union.