On the Computational Practicality of Private Information Retrieval
On the Computational Practicality of Private Information Retrieval
复制标题
DOI：
--
复制
发表时间：
2006
期刊：
影响因子：
--
通讯作者：
R. Sion;Bogdan Carbunar
中科院分区：
文献类型：
--
作者：
R. Sion;Bogdan Carbunar
我们探索单服务器计算私有信息检索（PIR）的局限性，以防止客户端访问模式泄漏。我们表明，在最近的真实硬件上部署非平凡的单服务器 PIR 协议的时间效率比简单地传输整个数据库要低几个数量级。我们强调，这些结果超出了现有知识的范围，即在不利的假设下“不切实际”。它们反映了现代硬件的固有局限性，可能是以通信成本为中心的协议设计的结果。我们认为，在可预见的未来，这可能会在非专业传统硬件上持续存在。我们在现代现成硬件上的实验设置中验证了我们的推理。最终，我们希望我们的结果能够激发实际设计。
We explore the limits of single-server computational private information retrieval (PIR) for the purpose of preserving client access patterns leakage. We show that deployment of non-trivial single server PIR protocols on real hardware of the recent past would have been orders of magnitude less time-efficient than trivially transferring the entire database. We stress that these results are beyond existing knowledge of mere “impracticality” under unfavorable assumptions. They rather reflect an inherent limitation with respect to modern hardware, likely the result of a communication-cost centric protocol design. We argue that this is likely to hold on non-specialized traditional hardware in the foreseeable future. We validate our reasoning in an experimental setup on modern off-the-shelf hardware. Ultimately, we hope our results will stimulate practical designs.