Privacy Without Losing Place: A Paradigm for Private Retrieval in Spatial RAGs

2026-05-06

AI Summary

This work addresses the challenge of balancing user location privacy and retrieval performance in spatial retrieval-augmented generation (Spatial RAG). To this end, it proposes a Privacy-preserving Anchor Substitution (PAS) mechanism that introduces structured relative positional encoding—replacing raw coordinates with anchors, directional intervals, and distance intervals—thereby achieving privacy protection without perturbing the original coordinates. The study reveals a non-monotonic relationship between privacy and utility and identifies a geometric bias effect induced by anchor discretization. Experimental results on synthetic urban datasets demonstrate that PAS achieves adversarial localization errors of approximately 370–400 meters while preserving over 50% of baseline retrieval performance, with downstream large language model generation quality remaining robust.

Abstract

This work introduces PAS -- Privacy Anchor Substitution, a structured mechanism for enabling user location privacy in spatial retrieval-augmented generation (RAG) systems. Unlike conventional differential privacy methods that directly perturb user locations, PAS represents location with relative anchor encoding consisting of an anchor, direction bin, and distance bin, allowing seamless integration with modern RAG pipelines. We evaluate PAS on a synthetic urban dataset and show that it achieves impressive coarse privacy guarantees, with approximately 370-400m adversarial location error, while retaining more than half of the baseline retrieval performance. Despite the slight drop in retrieval performance, the downstream generation quality under PAS remains comparatively robust, indicating that large language models can compensate for imperfect spatial retrieval. Furthermore, we provide empirical analysis showing that PAS exhibits non-monotonic privacy-utility relationship with respect to privacy parameters. We attribute this to geometric bias induced by anchor discretization, making it different from continuous noise mechanisms such as geo-indistinguishability. Our results show that structured spatial representations offer a practical approach to privacy in location based reasoning in RAG systems.

Problem

• location privacy
• spatial retrieval-augmented generation
• privacy-utility trade-off
• spatial representation
• RAG systems

Innovation

• Privacy Anchor Substitution
• Spatial RAG
• Location Privacy
• Relative Anchor Encoding
• Non-monotonic Privacy-Utility Trade-off