Research map / Adversarial robustness and security / Attacks on text and graphs
Local-Global Defense against Unsupervised Adversarial Attacks on Graphs
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 |
|---|---|---|---|
| combine local sensitive-edge defense with global defense | unsupervised graph representation learning · Cora, Citeseer, and Polblogs under PGD attacks | node classification accuracy (%) | Helped |
| add local defense with attention penalties on perturbed edges (ablation) | Ours · Cora, PGD 0.9 | node classification accuracy (%): 58.8±1.2 -> 66.2±1.4 | Helped |
| add local defense with attention penalties on perturbed edges (ablation) | Ours · Citeseer, PGD 0.9 | node classification accuracy (%): 42.8±1.0 -> 51.1±1.1 | Helped |
| add local defense with attention penalties on perturbed edges (ablation) | Ours · Polblogs, PGD 0.9 | node classification accuracy (%): 78.6±0.7 -> 83.3±0.6 | Helped |
| add local defense with attention penalties on perturbed edges (ablation) | Ours · Cora, PGD 0.9 | node clustering NMI (%): 24.7±1.1 -> 42.1±3.9 | Helped |
| add local defense with attention penalties on perturbed edges (ablation) | Ours · Citeseer, PGD 0.9 | node clustering NMI (%): 15.5±1.3 -> 20.8±4.3 | Helped |
| add local defense with attention penalties on perturbed edges (ablation) | Ours · Polblogs, PGD 0.9 | node clustering NMI (%): 29.8±1.7 -> 35.6±1.2 | Helped |
| use only one defense strategy instead of combining local and global defense (ablation) | Ours · Polblogs, PGD perturbation at 20% | node classification accuracy (%) | Hurt |
Closest papers by meaning
- Value at Adversarial Risk: A Graph Defense Strategy against Cost-Aware Attacks (AAAI 2024 · similarity 0.90)
- When Witnesses Defend: A Witness Graph Topological Layer for Adversarial Graph Learning (AAAI 2025 · similarity 0.88)
- If You Want to Be Robust, Be Wary of Initialization (NeurIPS 2024 · similarity 0.88)
- Adversarial Contrastive Graph Masked AutoEncoder Against Graph Structure and Feature Dual Attacks (AAAI 2025 · similarity 0.87)
- Adversarial Robust Generalization of Graph Neural Networks (ICML 2025 · similarity 0.87)
- Bounding the Expected Robustness of Graph Neural Networks Subject to Node Feature Attacks (ICLR 2024 · similarity 0.87)
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.