Datasets:

elsiddik
/

zz

license: apache-2.0
task_categories:
  - text-generation
  - text2text-generation
language:
  - en
  - code
tags:
  - cybersecurity
  - vulnerability-detection
  - cve
  - code-audit
  - code-repair
size_categories:
  - 1M<n<10M

FineSec-1M Master Security Dataset

FineSec-1M is a high-precision, 1.11 GB cybersecurity dataset containing 1,000,000 training examples (500,000 vulnerable code snippets + 500,000 safe control examples) designed for fine-tuning state-of-the-art security LLMs.

Dataset Structure

Server-Side Request Forgery (SSRF - CWE-918): 167,054 examples
Unquoted Search Paths (CWE-428): 166,525 examples
Insecure Randomness (CWE-330): 166,421 examples
Safe Control Code Suite (N/A): 500,000 examples

Class & Language Balance

Class Balance: Exactly 50.0% Vulnerable | 50.0% Safe Controls (Guarantees zero false positives).
Languages: Python (66.7%) & C/C++ (33.3%).

Usage with Hugging Face Datasets

from datasets import load_dataset

dataset = load_dataset('elsiddik/zz', split='train')
print(f'Total records: {len(dataset)}')