Research map / Adversarial robustness and security / Attacks on text and graphs
Generic Adversarial Attack Framework Against Graph-based Vertical Federated Learning
Yimin Liu, Peng Jiang, Qi Liu, Liehuang Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add generalized embedding learning (GEL) (ablation) | SGAC · Ali-CVR | attack success rate (%): 40.67 -> 53.50 | Improved |
| add domain knowledge transfer (DKT) on top of GEL (ablation) | SGAC · Ali-CVR | attack success rate (%): 53.50 -> 72.34 | Improved |
| add shadow input construction (SIC) on top of GEL and DKT (ablation) | SGAC · Ali-CVR | attack success rate (%): 72.34 -> 85.55 | Improved |
| add node-level saliency masking (ablation) | SGAC · Ali-CVR | attack success rate (%): 11.67±1.05 drop when removed | Improved |
| add edge-level saliency masking (ablation) | SGAC · Ali-CVR | attack success rate (%): 12.15±0.67 drop when removed | Improved |
| randomly drop edges instead of saliency-based masking (ablation) | SGAC · Ali-CVR | attack success rate (%): 10.46±1.52 decline | Hurt |
| add same-class shadow additions (ablation) | SGAC · Ali-CVR | attack success rate (%): 0.64±0.11 gain | No clear effect |
| increase auxiliary labeled nodes per class from 20 to 120 | SGAC · Ali-CVR | attack success rate (%): 76.54 -> 88.54 | Improved |
| increase benign passive parties from 2 to 32 | SGAC · Ali-CVR | attack success rate (%): 82.57 -> 75.27 | Hurt |
| use SGAC instead of Hijack surrogate attack | Hijack · Ali-CVR | attack success rate (%): 55.18 -> 85.60 | Improved |
| apply gradient filtering with top fraction 0.3 or lower | defense against SGAC · Ali-CVR | ASR and clean accuracy: clean accuracy 60.47% | Mixed |
| train server with adversarial training (unseen SGAC variant) | defense against SGAC · Ali-CVR | attack success rate (%): over 81.17% ASR remains | No clear effect |
Closest papers by meaning
- Value at Adversarial Risk: A Graph Defense Strategy against Cost-Aware Attacks (AAAI 2024 · similarity 0.86)
- Constructing Adversarial Examples for Vertical Federated Learning: Optimal Client Corruption through Multi-Armed Bandit (ICLR 2024 · similarity 0.86)
- Virtual Nodes Can Help: Tackling Distribution Shifts in Federated Graph Learning (AAAI 2025 · similarity 0.85)
- Label-Free Backdoor Attacks in Vertical Federated Learning (AAAI 2025 · similarity 0.85)
- VFLAIR: A Research Library and Benchmark for Vertical Federated Learning (ICLR 2024 · similarity 0.85)
- TRACI: A Data-centric Approach for Multi-Domain Generalization on Graphs (AAAI 2025 · similarity 0.84)
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.
