Scholarlink

In recent years, Private Information Retrieval (PIR) schemes capable of rapid response have gotten significant attention, with SimplePIR, for instance, achieving a throughput of approximately 10 GB/s in a single-threaded setting. However, these schemes often require users to download data related to the database, known as client hints, making them unsuitable for scenarios where the database undergoes frequent updates. This paper, building upon the Spiral scheme, incorporates the precomputation approach of SimplePIR to design a query method that leverages simulatable homomorphic ciphertexts for offline precomputation. This enhancement not only improves the efficiency of the Spiral scheme but also addresses the issue of SimplePIR’s necessity for downloading client hints. Experimental results demonstrate that, on a database with \(2^{20}\) entries of 256 bytes each, the proposed scheme achieves a throughput 1.87 times higher than that of the original Spiral scheme.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced PIR Scheme Combining SimplePIR and Spiral: Achieving Higher Throughput Without Client Hints

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Haoyao Xu,
  • Yitong Li,
  • Haibo Tian

摘要

In recent years, Private Information Retrieval (PIR) schemes capable of rapid response have gotten significant attention, with SimplePIR, for instance, achieving a throughput of approximately 10 GB/s in a single-threaded setting. However, these schemes often require users to download data related to the database, known as client hints, making them unsuitable for scenarios where the database undergoes frequent updates. This paper, building upon the Spiral scheme, incorporates the precomputation approach of SimplePIR to design a query method that leverages simulatable homomorphic ciphertexts for offline precomputation. This enhancement not only improves the efficiency of the Spiral scheme but also addresses the issue of SimplePIR’s necessity for downloading client hints. Experimental results demonstrate that, on a database with \(2^{20}\) entries of 256 bytes each, the proposed scheme achieves a throughput 1.87 times higher than that of the original Spiral scheme.