VIPIR: A Versatile GPU Framework for Integrating Private Information Retrieval Protocols

2026-06-09

AI Summary

This work addresses the bottlenecks of private information retrieval (PIR)—including high computational throughput demands, substantial memory usage, and significant bandwidth overhead—while preserving access pattern privacy. The authors propose a co-designed PIR protocol accelerated by GPUs, introducing a unified analytical framework that synergistically combines the strengths of two dominant PIR paradigms. A key innovation is the ExpPack (Expanded Ring Packing) compression technique, which enables efficient data representation. For the first time, tensor cores are leveraged to perform mixed-precision integer matrix multiplication within PIR, integrated with Number Theoretic Transform (NTT), optimized GEMM routines, and memory-efficient scheduling to achieve scalable multi-GPU execution. The resulting system improves throughput by multiple orders of magnitude over state-of-the-art approaches while substantially reducing communication costs and memory footprint, thereby enabling practical large-scale PIR deployment.

Abstract

While private information retrieval (PIR) enables private database services by fully concealing access patterns, it simultaneously requires high computational throughput, large memory capacity, and substantial memory bandwidth. We introduce VIPIR, a versatile GPU framework that co-designs PIR protocols with GPU acceleration. We develop a unified analytic model showing that state-of-the-art PIR protocols fall into two categories with complementary limitations, and propose two protocols that flexibly combine techniques across these categories, overcoming the limitations of both classes. These protocols incorporate a GPU-friendly data compression method called expansion-based ring packing (ExpPack), which offers a high degree of parallelism and minimal communication cost. VIPIR applies further optimizations to core operations, including number-theoretic transforms (NTTs) and various matrix-matrix multiplications (GEMMs). Notably, we develop a tensor-core-based execution method for database multiplication by interpreting it as a mixed-integer-type GEMM. We also design memory-efficient scheduling methods that minimize intermediate buffers and enable multi-GPU scaling under memory capacity constraints. Overall, VIPIR achieves orders-of-magnitude higher throughput than prior PIR systems while reducing communication and memory overheads, making large-scale PIR practical.

Problem

• Private Information Retrieval
• GPU acceleration
• Memory bandwidth
• Computational throughput
• Scalability

Innovation

• GPU acceleration
• Private Information Retrieval (PIR)
• Expansion-based Ring Packing (ExpPack)
• Tensor Core GEMM
• Memory-efficient Scheduling