metadata

dataset_info:
  features:
    - name: project
      dtype: string
    - name: commit_id
      dtype: string
    - name: func
      dtype: string
    - name: vul
      dtype: int64
    - name: CVE ID
      dtype: string
    - name: CWE ID
      dtype: string
    - name: CWE Name
      dtype: string
    - name: CWE Description
      dtype: string
    - name: Potential Mitigation
      dtype: string
    - name: __index_level_0__
      dtype: int64
  splits:
    - name: train
      num_bytes: 64041861
      num_examples: 15770
    - name: validation
      num_bytes: 9198025
      num_examples: 2253
    - name: test
      num_bytes: 17932135
      num_examples: 4506
  download_size: 25107430
  dataset_size: 91172021
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

CWE Enriched and Balanced BigVul+PrimeVul

This dataset is a processed and enriched version of the BigVul and PrimeVul datasets, enhanced using CWE mappings from MITRE's CWE Database. The purpose of this dataset is to provide a cleaner, more balanced, and more informative version of the original datasets for vulnerability prediction and CWE classification tasks.

📌 Dataset Description

This dataset contains vulnerability information from BigVul and PrimeVul datasets, with additional columns added from CWE ID mappings fetched from MITRE’s CWE database. The dataset was cleaned, balanced, and split into training, validation, and test sets while maintaining:

The distribution ratio of vulnerable and non-vulnerable functions.
The distribution ratio of each CWE ID across all splits.

💡 Key Features:

Project & Commit Information: Metadata related to the project and its commits.
Function Code (func): The code snippet associated with each entry.
Vulnerability Label (vul): Whether the function is vulnerable (1) or not (0).
CVE ID & CWE ID Mapping: Relevant IDs for categorizing vulnerabilities.
CWE Information: Additional information fetched from MITRE's CWE database including:
- CWE Name
- CWE Description
- Potential Mitigation

📂 Dataset Structure

The dataset consists of the following columns:

Column Name	Description
project	The name of the project where the code is taken from.
commit_id	The unique commit identifier.
func	The function code snippet.
vul	Label indicating if the function is vulnerable (`1`) or not (`0`).
CVE ID	The CVE identifier (if applicable).
CWE ID	The CWE identifier corresponding to the vulnerability.
CWE Name	The name of the CWE ID.
CWE Description	Description of the CWE category.
Potential Mitigation	Recommended mitigations for addressing the vulnerability.

📊 CWE IDs Contained In This Dataset

The dataset contains the following 13 unique CWE IDs (Including Not Applicable):

CWE ID	CWE Name
CWE-119	Improper Restriction of Operations within the Bounds of a Memory Buffer
CWE-20	Improper Input Validation
CWE-125	Out-of-bounds Read
CWE-399	Resource Management Errors
CWE-200	Information Exposure
CWE-787	Out-of-bounds Write
CWE-264	Permissions, Privileges, and Access Control
CWE-416	Use After Free
CWE-476	NULL Pointer Dereference
CWE-190	Integer Overflow or Wraparound
CWE-189	Numeric Errors
CWE-362	Concurrent Execution using Shared Resource with Improper Synchronization (Race Condition)
NOT_APPLICABLE	Not applicable / No CWE ID assigned

🔍 Data Splitting & Balancing

The dataset has been split into training, validation, and test sets using stratified sampling to ensure the distribution of:

Vulnerable (vul = 1) and Non-Vulnerable (vul = 0) functions.
Each CWE ID remains consistent across splits.

📥 Usage

To load this dataset from the Hugging Face Hub:

from datasets import load_dataset

dataset = load_dataset('mahdin70/cwe_enriched_balanced_bigvul_primevul')

📌 Citation

If you use this dataset, please cite the original sources of BigVul and PrimeVul datasets.