Datasets:

Modalities:
Text
Formats:
json
Languages:
English
ArXiv:
Libraries:
Datasets
Dask
License:
AZERG-Dataset / README.md
lekssays's picture
Update README.md
3e25de1 verified
---
license: apache-2.0
tags:
- threat intelligence
- cyber security
- STIX standard
- MITRE ATT&CK
task_categories:
- text-generation
language:
- en
size_categories:
- 10K<n<100K
---
# AZERG-Dataset
This repository contains the AZERG-Dataset, a comprehensive collection of annotated cyber threat intelligence (CTI) reports designed for training and evaluating models on STIX entity and relationship extraction.
This dataset was created for the paper: "From Text to Actionable Intelligence: Automating STIX Entity and Relationship Extraction". It is the largest publicly available dataset of its kind, meticulously annotated with STIX-compliant entities and relationships to facilitate the development of automated threat intelligence tools.
# πŸ“– Dataset Overview
The AZERG-Dataset is built from 141 real-world threat analysis reports and contains 4,011 STIX entities and 2,075 STIX relationships. It was curated to address the lack of training data for automated STIX report generation and supports a multi-task approach to threat intelligence extraction.
The extraction process is divided into four sequential subtasks:
- T1: Entity Detection: Identifying all STIX entities (SDOs and SCOs) in a text passage.
- T2: Entity Type Identification: Assigning a specific STIX type to each detected entity.
- T3: Related Pair Detection: Identifying which pairs of entities are semantically related based on the text.
- T4: Relationship Type Identification: Determining the precise STIX relationship type (e.g., uses, targets) between a related pair of entities.
## πŸ“‚ Dataset Structure
The dataset is organized into train and test splits. The training and testing data are sourced from completely non-overlapping reports and vendors to ensure a robust evaluation of model generalization.
```
AZERG-Dataset/
β”œβ”€β”€ train/
β”‚ β”œβ”€β”€ azerg_T1_train.json
β”‚ β”œβ”€β”€ azerg_T2_train.json
β”‚ β”œβ”€β”€ azerg_T3_train.json
β”‚ β”œβ”€β”€ azerg_T4_train.json
β”‚ └── azerg_MixTask_train.json # Combined data for all tasks
└── test/
β”œβ”€β”€ annoctr_T1_test.json
β”œβ”€β”€ annoctr_T2_test.json
β”œβ”€β”€ annoctr_T3_test.json
β”œβ”€β”€ annoctr_T4_test.json
β”œβ”€β”€ azerg_T1_test.json
β”œβ”€β”€ azerg_T2_test.json
β”œβ”€β”€ azerg_T3_test.json
└── azerg_T4_test.json
```
## πŸ“œ Citation
If you use this dataset in your research, please cite the original paper (ArXiv for now, the paper is accepted at RAID 2025):
```
@article{lekssays2025azerg,
title={From Text to Actionable Intelligence: Automating STIX Entity and Relationship Extraction},
author={Lekssays, Ahmed and Sencar, Husrev Taha and Yu, Ting},
journal={arXiv preprint arXiv:2507.16576},
year={2025}
}
```