Reflex: Speeding Up SMPC Query Execution through Efficient and Flexible Intermediate Result Size Trimming

2025-03-26

AI Summary

Secure multi-party computation (SMPC) suffers from high query execution overhead and intermediate result explosion, making it challenging to simultaneously achieve security and performance. Method: This paper proposes a fine-grained, pluggable intermediate result pruning mechanism. It generalizes pruning— for the first time—as an operator-level configurable and query-level customizable technique, achieved by extending MPC protocols, designing a standardized pruning interface, constructing a security-performance joint optimization model, and incorporating state-preserving optimizations. Pruning is dynamically adjusted under bounded, controllable information leakage constraints. Contribution/Results: Experiments demonstrate that our approach significantly improves query throughput over state-of-the-art methods while enhancing security guarantees. It provides a scalable, foundational building block for SMPC query planners, enabling efficient and secure query execution without compromising correctness or confidentiality.

Technology Category

Search and Optimization: Distributed SearchData Mining & Knowledge Management: Intelligent Query ProcessingMachine Learning: Privacy

Application Category

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

Abstract

There is growing interest in Secure Analytics, but fully oblivious query execution in Secure Multi-Party Computation (MPC) settings is often prohibitively expensive. Recent related works propose different approaches to trimming the size of intermediate results between query operators, resulting in significant speedups at the cost of some information leakage. In this work, we generalize these ideas into a method of flexible and efficient trimming of operator outputs that can be added to MPC operators easily. This allows for precisely controlling the security/performance trade-off on a per-operator and per-query basis. We demonstrate that our work is practical by porting a state-of-the-art trimming approach to it, resulting in a faster runtime and increased security. Our work lays down the foundation for a future MPC query planner that can pick different performance and security targets when composing physical query plans.

Problem

• Speeding up SMPC query execution efficiently
• Balancing security and performance trade-offs flexibly
• Enabling adaptable MPC query planning securely

Innovation

• Flexible trimming of MPC operator outputs
• Precise control of security-performance trade-off
• Faster runtime with increased security