Graph D2A: D2A Dataset for Vulnerability Detection Transformed into ECPG Format for Training GNNs

The Graph D2A is based on the dataset 'D2A: A Dataset Built for AI-Based Vulnerability Detection Methods Using Differential Analysis'. Graph D2A was created by transforming individual samples from D2A into the form of Extended Code Property Graphs (ECPGs). The resulting dataset is designed for graph neural network models (GNNs) for detection on vulnerabilities in source code.

Graph D2A was developed as part of the Master's thesis 'Evaluating Reliability of Static Analysis Results Using Machine Learning' at BUT FIT.