The first release of Graph Robustness Benchmark (GRB).
- API based on pure PyTorch, CogDL, and DGL.
- Include five graph datasets of different scales.
- Support graph injection attacks (e.g., RND, FGSM, PGS, SPEIT, TDGIA).
- Support adversarial defenses (e.g., layer normalization, adversarial training, GNNSVD, GNNGuard).
- Provide homepage.
- Provide leaderboards of all datasets.
- Provide basic documentation.
- Provide scripts for reproducing results.