Research map / Graph learning / GNN architectures
Imprint of the Forgotten: Stealthy Membership Inference in Unlearned Graph Neural Networks
He Zhang, Bang Wu, Xiaoning Liu, Karin Verspoor, Xun Yi
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 |
|---|---|---|---|
| use dual-encoder architecture with embedding loss instead of a plain MLP attack model (ablation) | FMIA · Cora/CS/DBLP/PubMed x GCN/GAT/GIN | attack AUC | Improved |
| remove encoder modules or swap embedding distance function (ablation) | FMIA · PubMed GNNDelete | attack AUC | Improved |
| increase amount of unlearned data under Fine-tuning (ablation) | unlearned GNN vulnerability · Fine-tuning unlearning | attack AUC | Hurt |
| increase amount of unlearned data under GIF (ablation) | unlearned GNN vulnerability · GIF unlearning | attack AUC | No clear effect |
| increase amount of unlearned data under GNNDelete (ablation) | unlearned GNN vulnerability · GNNDelete unlearning | attack AUC | Improved |
| use deeper DeepGCN (56 layers vs 3) as target (ablation) | unlearned GNN vulnerability · PubMed, Fine-tuning | attack AUC (%): 96.18 -> 93.44 | Hurt |
| use attack model trained with a different unlearning method (GIF -> Fine-tuning) (ablation) | FMIA · Fine-tuning target | attack AUC: 13.3% | Hurt |
| use attack model trained on a different dataset or architecture (transfer) (ablation) | FMIA · cross dataset and model | attack AUC | Mixed |
Closest papers by meaning
- A Cognac Shot To Forget Bad Memories: Corrective Unlearning for Graph Neural Networks (ICML 2025 · similarity 0.89)
- Erase Then Rectify: A Training-Free Parameter Editing Approach for Cost-Effective Graph Unlearning (AAAI 2025 · similarity 0.88)
- Community-Centric Graph Unlearning (AAAI 2025 · similarity 0.88)
- Adversarial Machine Unlearning (ICLR 2025 · similarity 0.88)
- Can Graph Neural Networks Expose Training Data Properties? An Efficient Risk Assessment Approach (NeurIPS 2024 · similarity 0.87)
- Prompt-based Unifying Inference Attack on Graph Neural Networks (AAAI 2025 · 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.
