VertiSplitRAG: Privacy-Preserving Retrieval-Augmented Generation Through Vertical Attribute Partitioning
Abstract
Retrieval-Augmented Generation systems enhance Large Language Model responses by incorporating external knowledge but inherently expose sensitive information through joint attribute encoding in embeddings. Existing privacy-preserving approaches apply post-hoc protection after embeddings are formed, leaving multi-attribute leakage vulnerabilities unaddressed. This thesis introduces VertiSplitRAG, a privacy-by-design framework that restructures data through vertical attribute partitioning before embedding, ensuring that no single component observes complete records. We evaluated four architectural variants: Baseline RAG (BAS), VertiSplitRAG for single-attribute queries (VSS), VertiSplitRAG for multi-attribute with shared LLM (VSM), and VertiSplitRAG with multi-LLM architecture (VSML), on synthetic medical patient records to assess privacy protection, answer quality, and system latency. Our multi-LLM architecture achieves complete cross-attribute isolation through information-theoretic structural separation, maintaining full privacy budget across all queries while other systems progressively exhaust their isolation guarantees over time. This privacy improvement comes at an acceptable utility cost, with privacy-preserving variants maintaining competitive faithfulness scores comparable to the baseline across query sequences. Additionally, the multi-LLM architecture demonstrates lower latency than the baseline for single-attribute queries, supporting practical deployment in…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.