Privacy-Preserving Near Neighbor Search via Sparse Coding with\n Ambiguation

2021/02/08 by Behrooz Razeghi, Razeghi, Behrooz, Sohrab Ferdowsi +7
Computer Science · Engineering · #Advanced Image and Video Retrieval Techniques #Automated Road and Building Extraction #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Remote-Sensing Image Classification

paper · pdf · doi:10.48550/arxiv.2102.04274

openalex publication_date 2021/02/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

In this paper, we propose a framework for privacy-preserving approximate near\nneighbor search via stochastic sparsifying encoding. The core of the framework\nrelies on sparse coding with ambiguation (SCA) mechanism that introduces the\nnotion of inherent shared secrecy based on the support intersection of sparse\ncodes. This approach is `fairness-aware', in the sense that any point in the\nneighborhood has an equiprobable chance to be chosen. Our approach can be\napplied to raw data, latent representation of autoencoders, and aggregated\nlocal descriptors. The proposed method is tested on both synthetic i.i.d data\nand real large-scale image databases.\n

Related

• Privacy Preserving Identification Using Sparse Approximation with Ambiguization
• Sub-Linear Privacy-Preserving Near-Neighbor Search
• Privacy-Preserving Identification via Layered Sparse Code Design:\n Distributed Servers and Multiple Access Authorization
• Reconstruction of Privacy-Sensitive Data from Protected Templates
• Differentially Private High-Dimensional Approximate Range Counting, Revisited