Search papers, labs, and topics across Lattice.
This paper introduces a novel approach to private dense retrieval that balances the need for confidentiality with retrieval efficiency by utilizing learned deep hashing as a private filter. By generating a randomized binary code, the method quickly narrows down the candidate documents to a shortlist, which is then reranked using encrypted techniques to ensure privacy. The results show that this approach achieves retrieval quality comparable to full-corpus methods while significantly reducing processing time and maintaining strong privacy guarantees across multiple datasets.
A learned shortlist can make private dense retrieval both accurate and practical, achieving near-full-corpus quality with a fraction of the computational cost.
Hosted retrieval-augmented generation (RAG) and semantic search allow users to query valuable provider-held corpora, raising two competing demands: to hide each query and chosen result, yet reveal only the documents that the user is authorized to receive. Existing cryptographic approaches either make this costly by processing the entire corpus for every query, or sacrifice quality for efficiency by scanning a few clusters. We repurpose learned deep hashing as a private filter: a randomized binary code points the provider to a short candidate list, while encrypted reranking and oblivious key transfer protect the precise query and final selection. This shortlist short-circuits full-corpus cryptographic search without sacrificing retrieval quality: with 200-500 candidates, it closely matches full-corpus retrieval across five zero-shot corpora spanning 25K to 5.4M documents. On the full 2.68M-passage NQ corpus over a 10-Gbps link, our protocol only adds 0.73 seconds, or 10 percent, to a 128-token Qwen3-32B RAG pipeline. The released code satisfies directional metric differential privacy (DP) and substantially reduces embedding-inversion and property-inference leakage, demonstrating that a carefully learned shortlist can make private dense retrieval both accurate and practical.
