Sauron Decompiler

---

Sauron Decompiler is a machine learning-powered project dedicated to the challenging task of decompiling binary executables back into human-readable C source code. Like the mythical phoenix rising from ashes, this project aims to reconstruct the original form and logic from its compiled representation.

---

🚀 Features

• Exhaustive EDA: Deep analysis of the ExeBench dataset, including patterns, distributions, and insights within binary code and C source.
• Robust Data Cleaning: Pipelines to ensure dataset quality and integrity for effective ML model training.
• Decompilation Objective: Train and evaluate models to translate binary code snippets into semantically equivalent C code.
• Web Interface: Django-based webapp for interactive decompilation and result visualization.

📦 Repository Structure

• EDA/ — Exploratory Data Analysis scripts and visualizations
• fine_tuning/ — Model training, evaluation, and fine-tuning scripts
• preprocessing/ — Data cleaning and preparation
• test_finetuning/ — Test scripts and results for fine-tuned models
• web/ — Django webapp for interactive decompilation
• data/ — Datasets and processed data
• doc/ — Documentation and assets

🛠️ Quickstart

1. Clone the repository

git clone https://github.com/archibald-carrion/decompiler.git
cd decompiler

2. Install dependencies

When preprocesing:

pip install -r preprocessing/requirements.txt

When running demo:

pip install -r fine_tuning/requirements.txt
pip install -r web/requirements.txt

3. Set up environment variables

The Django webapp requires a secret key for security. Set it before running the server:

On Windows (cmd.exe):

set DJANGO_SECRET_KEY=your-very-secret-key

On Linux/macOS:

export DJANGO_SECRET_KEY=your-very-secret-key

4. Run the Django webapp

cd web
python manage.py migrate
python manage.py runserver

The app will be available at http://127.0.0.1:8000/decompile/

⚙️ Flags & Configuration

• DJANGO_SECRET_KEY (required): Secret key for Django webapp security.
• DEBUG (in settings.py): Set to False in production.
• ALLOWED_HOSTS (in settings.py): Add your domain/IP for deployment.

📖 Documentation & Resources

• Django Documentation
• Python Official Docs
• CRISP-DM Methodology

🧠 Methodology

The project follows the CRISP-DM (Cross-Industry Standard Process for Data Mining) methodology:

graph TD;
    A[Business Understanding] --> B[Data Understanding];
    B --> C[Data Preparation];
    C --> D[Modeling];
    D --> E[Evaluation];
    E --> F[Deployment];
    F -->|Iterate| A;
    E -->|Refine Data| B;
    D -->|Optimize| C;