IM-PIR: In-Memory Private Information Retrieval

2025-09-08

AI Summary

Multi-server private information retrieval (PIR) suffers from prohibitive computational overhead and severe memory bandwidth bottlenecks—particularly during GB-scale database scans—limiting practical deployment. To address this, we propose the first PIM-based multi-server PIR architecture, offloading core PIR computations to UPMEM’s in-memory processing units. This design enables massive parallelism and near-data processing at the hardware level, eliminating costly data movement between memory and CPU prevalent in conventional implementations and thereby substantially alleviating memory bandwidth pressure. Experimental evaluation demonstrates over 3.7× higher query throughput compared to a state-of-the-art CPU-only implementation, while preserving information-theoretic security. To our knowledge, this is the first systematic integration of the PIM paradigm into multi-server PIR, establishing a novel architectural foundation for high-performance, privacy-preserving querying.

Technology Category

• Machine Learning: Hardware-aware ML
• Data Mining & Knowledge Management: Intelligent Query Processing
• Computer Vision: Multi-modal Vision

Application Category

• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

Private information retrieval (PIR) is a cryptographic primitive that allows a client to securely query one or multiple servers without revealing their specific interests. In spite of their strong security guarantees, current PIR constructions are computationally costly. Specifically, most PIR implementations are memory-bound due to the need to scan extensive databases (in the order of GB), making them inherently constrained by the limited memory bandwidth in traditional processor-centric computing architectures.Processing-in-memory (PIM) is an emerging computing paradigm that augments memory with compute capabilities, addressing the memory bandwidth bottleneck while simultaneously providing extensive parallelism.Recent research has demonstrated PIM's potential to significantly improve performance across a range of data-intensive workloads, including graph processing, genome analysis, and machine learning. In this work, we propose the first PIM-based architecture for multi-server PIR. We discuss the algorithmic foundations of the latter and show how its operations align with the core strengths of PIM architectures: extensive parallelism and high memory bandwidth. Based on this observation, we design and implement IM-PIR, a PIM-based multi-server PIR approach on top of UPMEM PIM, the first openly commercialized PIM architecture. Our evaluation demonstrates that a PIM-based multi-server PIR implementation significantly improves query throughput by more than 3.7x when compared to a standard CPU-based PIR approach.

Problem

• Addressing computational cost of private information retrieval
• Overcoming memory bandwidth limitations in PIR systems
• Leveraging processing-in-memory for multi-server PIR efficiency

Innovation

• Uses processing-in-memory for private information retrieval
• Leverages UPMEM PIM architecture for enhanced parallelism
• Achieves 3.7x higher throughput than CPU-based approaches