Datasets:

madox81
/

Adema_ATTACK_DS

dataset_info:
  features:
    - name: messages
      list:
        - name: content
          dtype: string
        - name: role
          dtype: string
  splits:
    - name: train
      num_bytes: 5886301
      num_examples: 21281
    - name: validation
      num_bytes: 723353
      num_examples: 2660
    - name: test
      num_bytes: 734972
      num_examples: 2661
  download_size: 2026961
  dataset_size: 7344626
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
license: apache-2.0
task_categories:
  - text-classification
  - text-generation
tags:
  - Cybersecurity

Adema ChatML Dataset

This dataset is a reformatted version of the original Adema research dataset, structured specifically in ChatML format for seamless integration with SFTTrainer and ChatML-based instruct models. Dataset has three splits:

Train: 80%
Validation: 10%
Test: 10%

Dataset Details

Original Paper: Automated discovery and mapping ATT&CK tactics and techniques for unstructured cyber threat intelligence
Format: ChatML (messages column containing role and content keys)
Task: Fine-tuning LLMs for cybersecurity applications.
Repository: Original Adema GitHub

Citation

@article{LI2024103815,
title = {Automated discovery and mapping ATT&CK tactics and techniques for unstructured cyber threat intelligence},
journal = {Computers & Security},
volume = {140},
pages = {103815},
year = {2024},
issn = {0167-4048},
doi = {https://doi.org/10.1016/j.cose.2024.103815},
url = {https://www.sciencedirect.com/science/article/pii/S0167404824001160},
author = {Lingzi Li and Cheng Huang and Junren Chen},
keywords = {Cyber threat intelligence, MITRE ATT&CK, Multi-label classification, Network security},
abstract = {As cyber attacks are growing, Cyber Threat Intelligence (CTI) enhances the ability of security systems to resist novel cyber threats. However, since most CTI is unstructured data written in natural language, it needs to be understood and summarized by security experts to be effectively utilized. To address the problem, we adopt the ATT&CK matrix as the taxonomy to propose a method for automated mapping of unstructured threat intelligence to tactics and techniques. The proposed method contains a pre-processor for text denoising, a label extractor for classifying which tactics and techniques category the text belongs to, and a post-processor for correcting the classification results. The label extractor consists of two multi-label classifiers based on DistilBERT for tactics and techniques classification respectively. The post-processor corrects the classification results based on the relations between tactics, techniques, and sub-techniques in the matrix, eliminating errors caused by the independence between categories. In the evaluation, we collect the text data from the ATT&CK knowledge base and real cyber threat reports to build an experiment dataset, which contains 26,602 sentence samples. We apply the proposed method to the dataset to verify its effectiveness. The results show that the proposed method can accurately retrieve tactics and techniques with F0.5 score of 85.50% and 75.17% respectively, which outperforms the baseline method by about 10%.}
}