---
license: mit
task_categories:
- text-retrieval
language:
- en
tags:
- cybersecurity
- benchmark
- retrieval
- mitre-attack
- sigma
- cve
- cwe
- soc
- evaluation
pretty_name: CyberSec Retrieval Benchmark
size_categories:
- 1K<n<10K
configs:
- config_name: queries
  data_files:
  - split: train
    path: queries/train-*
- config_name: corpus
  data_files:
  - split: train
    path: corpus/train-*
- config_name: qrels
  data_files:
  - split: train
    path: qrels/train-*
---

# CyberSec Retrieval Benchmark

Multi-task retrieval benchmark for cybersecurity embedding and reranking models.

## Configs

| Config | Contents |
|---|---|
| queries | Evaluation queries with 	ask labels |
| corpus | Document passages per task |
| qrels | Relevance judgments (
elevance=1 for gold docs) |

`python
from datasets import load_dataset
queries = load_dataset("alirezaaminzadeh/cybersec-retrieval-benchmark", "queries", split="train")
corpus = load_dataset("alirezaaminzadeh/cybersec-retrieval-benchmark", "corpus", split="train")
qrels = load_dataset("alirezaaminzadeh/cybersec-retrieval-benchmark", "qrels", split="train")
`

## Tasks

1. attack_retrieval
2. sigma_retrieval
3. cve_similarity
4. cwe_retrieval
5. threat_report_retrieval
6. soc_playbook

## Metrics

Recall@5, Recall@10, nDCG@10 per task and macro-average.