Software Vulnerability Prediction in Low-Resource Languages

This is the README file for the reproduction package of the paper: "Software Vulnerability Prediction in Low-Resource Languages: An Empirical Study of CodeBERT and ChatGPT".

I. Description

This project focuses on predicting software vulnerabilities using GPT-3.5-turbo model. It comprises data preprocessing, model training, and result analysis stages to predict vulnerabilities in software code.

II. Repository Structure

• DataPreprocessing.ipynb: A Jupyter notebook for reading the dataset and removing code comments to prepare the data for model training.
• OtherLanguageVulPrediction.ipynb: Implements a GPT model to predict software vulnerabilities. The model is trained using the cleaned dataset to detect and predict potential vulnerabilities.
• data/: Contains the original and pre-processed datasets, ready for use in model training and evaluation.
• utils/: Holds utility scripts and code snippets used in DataPreprocessing.ipynb and OtherLanguageVulPrediction.ipynb for data manipulation, preprocessing tasks, and API interactions.

III. Setup and Usage

Prerequisites:

• Python (3.x recommended)
• Jupyter Notebook or JupyterLab
• OpenAI API key for using GPT models

Steps

1. Install Dependencies Install all the required Python packages that list on OtherLanguageVulPrediction.ipynb

2. Set up your OpenAI API Key

• Obtain an API key from OpenAI's developer portal.
• In OtherLanguageVulPrediction.ipynb, replace openai.api_key = "" with your API key, ensuring the key is kept confidential.

Usage

1. Start Jupyter Notebook or JupyterLab: jupyter notebook

2. Open DataPreprocessing.ipynb and run the cells to preprocess the dataset.

3. Open OtherLanguageVulPrediction.ipynb to train the model and view the prediction results.