New PriDyG framework enhances privacy in dynamic graph inference using LLM-GNN collaboration

Researchers have developed PriDyG, a novel framework for privacy-preserving dynamic graph inference. This system combines graph neural networks (GNNs) with large language models (LLMs) to protect sensitive edge information in evolving relational datasets. PriDyG employs incremental private multi-hop aggregation, ensuring that the privacy cost remains constant regardless of the number of model updates, unlike traditional methods that accumulate privacy loss. Experiments demonstrate that PriDyG significantly outperforms existing baselines under identical privacy budgets, achieving utility comparable to naive retraining while drastically reducing cumulative privacy costs. AI

IMPACT Enhances privacy guarantees for dynamic graph analysis, potentially enabling more sensitive data to be used in machine learning applications.

RANK_REASON The cluster contains a research paper detailing a new method for privacy-preserving dynamic graph inference. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Edge-level Differentially Private Dynamic Graph Inference
• Gotit.pub
• graph neural network
• Hugging Face
• LLM
• PriDyG
• ScienceCast

• paper
• safety

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