Research map / Graph learning / GNN architectures
Bandits for Structure Perturbation-Based Black-Box Attacks To Graph Neural Networks With Theoretical Guarantees
Check 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 bandit feedback to guide structure perturbations instead of the RL attack | Black-box GNN attacks · Cora, Citeseer, Pubmed, MNIST, and CIFAR10 | attack success rate: 10%-20% higher | Helped |
| Use bandit feedback to guide structure perturbations instead of the RL attack | Black-box GNN attacks · Citation graphs and image graphs | queries for comparable attack performance: 1.5-3x queries | Helped |
| Increase the maximum number of perturbed edges | Bandit attack · Citation and image graphs | attack success rate | Helped |
| Increase the query budget | Bandit attack · Citation and image graphs | attack success rate | Helped |
| Increase the total cost budget | Bandit attack · GCN on citation graphs | attack success rate | Helped |
| Increase the cost budget after it is sufficient for the permitted edge changes (ablation) | Bandit attack · GCN on citation graphs | attack success rate | No clear effect |
| Increase projection scale α (ablation) | Bandit attack · Cora, Citeseer, Pubmed, MNIST, and CIFAR10 | attack success rate | Helped |
| Increase update step δ (ablation) | Bandit attack · Cora, Citeseer, Pubmed, MNIST, and CIFAR10 | attack success rate | Hurt |
| Vary α or δ across the tested values (ablation) | Bandit attack · Node and graph classification | attack success rate variation: no more than 10% | No clear effect |
Closest papers by meaning
- Highly Imperceptible Black-Box Graph Injection Attacks with Reinforcement Learning (AAAI 2025 · similarity 0.90)
- Deterministic Certification of Graph Neural Networks against Graph Poisoning Attacks with Arbitrary Perturbations (CVPR 2025 · similarity 0.90)
- A Simple and Yet Fairly Effective Defense for Graph Neural Networks (AAAI 2024 · similarity 0.88)
- Value at Adversarial Risk: A Graph Defense Strategy against Cost-Aware Attacks (AAAI 2024 · similarity 0.88)
- Stronger and Transferable Node Injection Attacks (AAAI 2024 · similarity 0.88)
- On the Interplay between Graph Structure and Learning Algorithms in Graph Neural Networks (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.