Pillar enables private and faster document search without revealing query details
PILLAR: Private Inverted-Index Lexical Lookup for Augmented Retrieval
Artificial Intelligence
Summary
When people search for information, usually the server hosting the documents sees the query, which can reveal private information. The authors introduce PILLAR, a system that lets a user search a public document collection privately, so the server doesn't know what the user searched for or which documents they accessed. It works by first using a private method to quickly find documents that share keywords with the query and then locally re-ranks these using dense representations to give better results. This approach reduces the time and quality problems seen in earlier private search systems.
What this means in practice
- •For search engine developers: Implement private document search systems that conceal user queries and access patterns without sacrificing speed or result quality.
- •For privacy tool builders: Create tools that allow users to query public corpora privately, enhancing information retrieval without exposing sensitive query information.
Authors
Truong Son Nguyen, Daniel Blackley, Ni Trieu, Evgenios M. Kornaropoulos
Abstract
Retrieval-augmented generation (RAG) hands the user's query to whoever hosts the corpus. We propose PILLAR, a Privacy-Preserving RAG (PPRAG) system based on Private Information Retrieval (PIR) in which a client utilizes the k documents most similar to their query from a server-held and publicly known corpus to respond to their query, while the server learns nothing about the query, either its terms or its access pattern. Prior PPRAG constructions rely on dense retrieval alone, translating approximate nearest-neighbor search into many query-dependent rounds of PIR, and pay for it in both latency and retrieval quality. PILLAR instead performs private hybrid retrieval in two stages. A sparse stage issues a small, fixed number of PIR queries against a carefully designed index of precomputed BM25 scores, filtering the corpus down to candidates that share terms with the query without the server ever seeing which terms these are. A dense stage then fetches only those candidates' document embeddings and re-ranks them locally, avoiding the many costly PIR queries that private dense retrieval typically requires. We instantiate PILLAR with two protocols that trade latency against retrieval quality, each built on a different private rendering of lexical search. PILLAR-Bin bins posting lists into a hash table and is a single-round design that achieves lower latency than state-of-the-art private retrieval schemes. PILLAR-Tree turns block-max pruning into an oblivious tree traversal combined with cuckoo hash tables and achieves the highest retrieval quality at lower latency than state-of-the-art schemes.