VulTypeRAG: Improving Software Vulnerability Type Identification via Retrieval-Augmented Generation and CWE Knowledge Graph

This is the source code to the paper "VulTypeRAG: Improving Software Vulnerability Type Identification via Retrieval-Augmented Generation and CWE Knowledge Graph". Please refer to the paper for the experimental details.

Approach

About dataset

The dataset files are large in size, so we have stored them in Google Drive: Google Drive Link.

About the experimental results in the paper:

1.The results for RQ1, RQ2, RQ3, RQ4, and RQ5 are stored in their corresponding folders.
2.The experimental results for the discussion section are stored in the discussion1 and discussion2 folders.

For reproducing the experiments:

1.Download the dataset and configure the file paths.
2.Run RQ2-RQ5.py to call the LLM and perform the SVTI task.
3.Run evaluation.py to conduct the metric evaluation.