Datasets:
language:
- en
license: mit
size_categories:
- 1K<n<10K
task_categories:
- token-classification
pretty_name: CyNER
dataset_info:
features:
- name: tokens
sequence: string
- name: ner_tags
sequence:
class_label:
names:
'0': B-Date
'1': B-Indicator
'2': B-Location
'3': B-Malware
'4': B-Organization
'5': B-System
'6': B-Threat_group
'7': B-Vulnerability
'8': I-Date
'9': I-Indicator
'10': I-Location
'11': I-Malware
'12': I-Organization
'13': I-System
'14': I-Threat_group
'15': I-Vulnerability
'16': O
splits:
- name: train
num_bytes: 3711236
num_examples: 7751
- name: validation
num_bytes: 796952
num_examples: 1661
- name: test
num_bytes: 784557
num_examples: 1662
download_size: 1579340
dataset_size: 5292745
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
CyNER 2.0 Augmented Dataset
CyNER 2.0 Augmented Dataset is a comprehensive dataset designed for Named Entity Recognition (NER) tasks, specifically tailored for cybersecurity-related text. This dataset combines an original cybersecurity NER dataset (bnsapa/cybersecurity-ner) with additional augmented data (taken from OPENCTI using AlienValut Connect), resulting in a robust resource for training and evaluating models on cybersecurity entity recognition.
Dataset Overview
The dataset is divided into three splits:
- Train: 7,751 examples
- Validation: 1,661 examples
- Test: 1,662 examples
Each example in the dataset consists of a sequence of tokens and their corresponding NER tags. The dataset has been carefully preprocessed to ensure high quality and consistency. The dataset mostly contains threat and intelligence report descriptions from the Alien Vault
Features
tokens: A sequence of strings representing the tokens in the text.
ner_tags: A sequence of strings representing each token's Named Entity Recognition tags.
NER Tags
The NER tags follow the BIO (Begin, Inside, Outside) tagging format and cover various cybersecurity-related entities, including but not limited to:
- B-Indicator: Indicators of Compromise (IOCs) such as IP addresses, file hashes, and domain names.
- I-Indicator: Continuation of an Indicator entity.
- B-Malware: Names of malware, trojans, or viruses.
- I-Malware: Continuation of a Malware entity.
- B-Organization: Names of organisations, companies, or groups.
- I-Organization: Continuation of an Organization entity.
- B-System: Operating systems, software, and platforms.
- I-System: Continuation of a System entity.
- B-Vulnerability: Known vulnerabilities or CVEs (Common Vulnerabilities and Exposures).
- I-Vulnerability: Continuation of a Vulnerability entity.
- B-ThreatActor: Threat actor names or groups (e.g., Threat_group).
- I-ThreatActor: Continuation of a ThreatActor entity.
- B-Date: Specific dates related to incidents or reports.
- I-Date: Continuation of a Date entity.
- O: Outside of any named entity.
Usage
This dataset is suitable for training, fine-tuning, and evaluating NER models, particularly those designed for cybersecurity applications. It has been structured to work seamlessly with popular NLP frameworks such as Hugging Face's transformers and dataset libraries.
License
This dataset is released under the MIT License, which allows for open and flexible usage with minimal restrictions. Users are free to use, modify, and distribute the dataset, provided that proper attribution is given to the original creators.
Citation
If you use this dataset in your research, please cite it as follows:
@dataset{pranava_kailash_2024_cyner,
title={CyNER 2.0 Augmented Dataset},
author={Pranava Kailash},
year={2024},
howpublished={Hugging Face, https://huggingface.co/datasets/PranavaKailash/CyNER2.0_augmented_dataset},
license={MIT}
}
Contact
For questions or issues regarding the dataset, please get in touch with pranavakailashsp@gmail.com.