Holo-NN proposes private nearest-neighbor search via compressive sensing
Original title: Holo-NN: Holographic Nearest Neighbor Search via Compressive Sensing
Encrypted queries reduce to one matrix-vector multiply, without FHE bootstrapping, interaction, or a client index.
The paper presents Holo-NN, an algebraic framework for privacy-preserving approximate nearest-neighbor search over high-dimensional embeddings. It replaces data-dependent branching used by indexes such as HNSW and ScaNN with a single linear projection and compressive sensing. The database is folded offline into a holographic sensing matrix, and encrypted search needs no bootstrapping or client-side index state. The authors compare it with other private ANN methods.
- Neighborhood membership is treated as sparse even when distances are dense.
- Query-time search is a non-interactive matrix-vector multiplication on CPUs or GPUs.
- The paper says accuracy is tunable and scales to production databases.
- Authors situate Holo-NN against other private approximate nearest-neighbor algorithms.
- Products
- Holo-NN
From the original article
1 min read at arxiv.orgIn the age of AI, privacy-preserving approximate nearest neighbor (ANN) search over high-dimensional embeddings has emerged as a critical yet frustratingly challenging domain, dominated by tradeoffs between speed, privacy, and accuracy. Traditional tree and graph-based indexing structures (e. g.
By Bryant Gipson · 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.