arXiv:2606. 26373v1 Announce Type: cross Abstract: Dense embeddings power semantic search and retrieval-augmented generation, but embedding-inversion attacks can reconstruct source text from a vector: when a vector database leaks, the documents behind it leak too.
By Sergey Kurilenko
The paper presents a practical private dense retrieval system that uses learned deep hashing as a private filter to generate a short candidate list for each query. Encrypted reranking and oblivious key transfer protect the exact query and final selection, allowing the system to match full‑corpus retrieval quality with only 200‑500 candidates. Experiments on five zero‑shot corpora and the 2.68M‑passage NQ corpus show minimal latency overhead and strong privacy guarantees.
By Peichun Hua, Danyang Chen, Junan Zhang, Haifeng Sun, Jingyu Wang, Diwen Xue, Mingyu Li, Yunming Xiao
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 onl...
arXiv:2609. 01945v1 Announce Type: cross Abstract: Federated Learning enables multiple clients to train a shared model while keeping their local datasets isolated.
By Miguel Morona-M\'inguez, Fernando P\'erez-Gonz\'alez, Alberto Pedrouzo-Ulloa
arXiv:2609.36376v1 Announce Type: cross
Abstract: Dense retrieval, the key component of Retrieval Augmented Generation (RAG), retrieves the most relevant documents by comparing dense vector represent...
By Louis Tremblay Thibault, Sofiane Azogagh, Marc-Olivier Killijian, Ulrich A\"ivodji
HEAT introduces a fine‑tuning method that treats the number of iterations used to approximate nonlinearities in fully homomorphic encryption (FHE) as learnable parameters, allowing them to co‑adapt with model weights. By optimizing iteration counts per nonlinearity, HEAT reduces the required iterations, bootstraps, and overall latency for encrypted GPT‑2 decoding while improving decode agreement. The approach achieves a 3.1× reduction in iterations, a 1.6× reduction in bootstraps, and a 1.4× speed‑up in end‑to‑end latency without changing the model architecture or requiring retraining from scratch.
By Alessandro Zirilli, Davide Marincione, Evgenios M. Kornaropoulos, Giuseppe Ateniese, Emanuele Rodol\`a