VulnCodeBERT

VulnCodeBERT is a PyTorch-based model for detecting vulnerabilities in code using the CodeBERT architecture. This project is inspired by the paper "Transformer-Based Language Models for Software Vulnerability Detection" and leverages the VulDeePecker dataset for training.

Overview

• Model Architecture: PyTorch-based CodeBERT
• Paper Reference: Transformer-Based Language Models for Software Vulnerability Detection
• Dataset Used: VulDeePecker
• Performance Metrics (Epoch 3):
  • Accuracy: 0.9623
  • Precision: 0.9538
  • Recall: 0.9526
  • F1 Score: 0.9532
  • False Positive Rate: 0.0311
  • False Negative Rate: 0.0474

Usage

Requirements

• Python 3.x
• PyTorch
• Transformers (HuggingFace)

License

This project is licensed under the MIT License.

Acknowledgments

• Transformer-Based Language Models for Software Vulnerability Detection