arXiv proposes differentially private framework for GraphRAG
Original title: Differentially Private GraphRAG: Protecting Relation Connectivity and Semantics
RelGuard is the first differentially private framework for GraphRAG providing formal privacy guarantees.
Researchers propose RelGuard, the first differentially private framework for Graph Retrieval-Augmented Generation. The framework mitigates connectivity leakage and protects relation semantics through relation removal. It formalizes relation differential privacy and uses a four-step sanitization mechanism before GraphRAG indexing.
- RelGuard provides formal privacy guarantees for GraphRAG
- Uses exponential mechanism for relation sanitization
- Experiments on three GraphRAG systems and two reconstruction attacks
- Improves protection over four existing defense mechanisms
- Preserves downstream utility better than alternatives
From the original article
1 min read at arxiv.orgGraph Retrieval-Augmented Generation (GraphRAG) connects information across documents through entity relationships to support relational reasoning. However, black-box access to GraphRAG systems can enable adversaries to reconstruct knowledge graphs and recover both structural connectivity between entities and the semantics of their relationships.
By Jie Fu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.