Research map / Graph learning / GNN architectures
Let Graph Be the Go Board: Gradient-Free Node Injection Attack for Graph Neural Networks via Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · Cora | misclassification rate (%): 29.5$_{\pm 2.8}$ -> 36.3$_{\pm 2.7}$ | Helped |
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · Citeseer | misclassification rate (%): 44.2$_{\pm 2.2}$ -> 49.4$_{\pm 2.8}$ | Helped |
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · Amazon Computers | misclassification rate (%): 39.9$_{\pm 1.3}$ -> 42.2$_{\pm 1.1}$ | Helped |
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · Amazon Photo | misclassification rate (%): 32.5$_{\pm 0.7}$ -> 33.6$_{\pm 1.2}$ | Helped |
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · OGB-Products | misclassification rate (%): 95.0$_{\pm 0.4}$ -> 97.4$_{\pm 0.4}$ | Helped |
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · Reddit | misclassification rate (%): 94.6$_{\pm 1.2}$ -> 98.7$_{\pm 0.6}$ | Helped |
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · Pubmed | misclassification rate (%): 68.3$_{\pm 1.0}$ -> 74.1$_{\pm 0.8}$ | Helped |
| Use gradient-free G²A2C instead of the strongest listed attacker | single-node injection against GCN · Wiki. CS | misclassification rate (%): 84.2$_{\pm 1.1}$ -> 86.6$_{\pm 0.8}$ | Helped |
| Use gradient-free G²A2C rather than the strongest listed baseline | attack with an architecture-mismatched attacker · Cora, Citeseer, and Pubmed across GCN, SGC, GAT, and APPNP | misclassification rate (%) | Helped |
| Increase the injected-node budget | G²A2C · GCN across datasets | misclassification rate (%): 3 | Helped |
| Increase the feature-shift budget | G²A2C · GCN across datasets | misclassification rate (%): 0.5 | Helped |
| Increase the edge budget | G²A2C · GCN across datasets | misclassification rate (%) | Helped |
Closest papers by meaning
- Highly Imperceptible Black-Box Graph Injection Attacks with Reinforcement Learning (AAAI 2025 · similarity 0.94)
- Stronger and Transferable Node Injection Attacks (AAAI 2024 · similarity 0.90)
- Grimm: A Plug-and-Play Perturbation Rectifier for Graph Neural Networks Defending Against Poisoning Attacks (AAAI 2025 · similarity 0.89)
- Value at Adversarial Risk: A Graph Defense Strategy against Cost-Aware Attacks (AAAI 2024 · similarity 0.88)
- Graph Agent Network: Empowering Nodes with Inference Capabilities for Adversarial Resilience (AAAI 2025 · similarity 0.88)
- CEGA: A Cost-Effective Approach for Graph-Based Model Extraction and Acquisition (ICML 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.