AI-powered secret detection for code commits using transformers and regex.

NeuroLeaks combines traditional regex scanning with CodeBERT, a state-of-the-art transformer model, to detect secrets like API keys, passwords, tokens, and credentials — even when they're obfuscated or renamed.

git clone https://github.com/YOUR_USERNAME/neuroleaks.git
cd neuroleaks

It's recommended to do this inside a virtual environment.

pip install -r requirements.txt

pip install pre-commit

repos:
  - repo: https://github.com/markorskip/neuroleaks
    rev: main  # or use a specific tag like v0.1.0
    hooks:
      - id: neuroleaks

From the root of your project:

pre-commit install

pre-commit run --all-files

You’ll get output like:

NeuroLeaks Secret Detection......................................❌
- [REGEX] Line 3: 'api_key = "AIzaSy123..."' matched api key pattern
- [ML]    Line 9: 'token = ghp_abcd1234...' (risk score: 0.91)

NeuroLeaks includes a test suite to validate its ML and regex behavior.

pytest tests/

• Regex Engine: Scans added lines for common secret patterns (e.g., AWS keys, GitHub tokens, passwords).
• Transformer Model: Uses microsoft/codebert-base to assess whether a line is likely to contain a secret, even if it's obfuscated or renamed.
• Hybrid Detection: If regex doesn’t catch it, ML does. Both methods are used on staged diffs via git diff --cached.

MIT License

Want to display that your repo is checked by NeuroLeaks? Add this badge:

![NeuroLeaks Verified](https://img.shields.io/badge/Secrets-Checked%20by%20NeuroLeaks-blueviolet?style=flat&logo=ai)

• Docker support
• Fine-tuned CodeBERT on real-world leaked secrets
• GitHub Action for CI integration
• Web dashboard to review and manage leaks

Open an issue or start a discussion. Contributions welcome!