🔒 AI-Powered Code Vulnerability Detection System

Deep learning system for detecting security vulnerabilities in source code using fine-tuned CodeT5.

🎯 Project Overview

Binary classification of source code: Safe vs Vulnerable

Features:

• ✅ CodeT5 transformer model (220M parameters)
• ✅ Fine-tuned on 10K balanced samples (optimized for RTX 3050)
• ✅ Mixed precision training (FP16)
• ✅ Interactive Streamlit UI
• ✅ 84-88% expected F1-score

💻 System Requirements

Hardware:

• GPU: NVIDIA RTX 3050 (4GB VRAM) or better
• RAM: 16GB
• Storage: 5GB free

Software:

• Python 3.8+
• CUDA 11.8+
• Windows 10/11

🚀 Quick Start

1. Installation

# Clone repository
git clone https://github.com/pranavyk10/code-vulnerability-detector.git
cd code-vulnerability-detector

# Create virtual environment
python -m venv .venv
.venv\Scripts\activate  # Windows
# source .venv/bin/activate  # Linux/Mac

# Install dependencies
pip install -r requirements.txt