New GNNBleed attack reveals private graph data

Researchers have developed a new method called GNNBleed to infer private edges in graph neural networks (GNNs). This attack works even with limited black-box access to the GNN model and is effective on dynamic graphs where the graph structure changes over time. GNNBleed significantly outperforms existing methods, achieving high F1 scores in both static and dynamic scenarios. AI

IMPACT This research highlights potential privacy vulnerabilities in GNNs, necessitating the development of more robust privacy-preserving techniques for graph data.

RANK_REASON The item is a research paper detailing a new attack method on graph neural networks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• GNNBleed
• Gotit.pub
• graph neural networks
• Hugging Face
• ScienceCast
• Zeyu Song

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →