FABLE - Batched Evaluation on Confidential Lookup Tables in 2PC

Learn about FABLE, a novel system for efficiently performing batched evaluation on confidential lookup tables in secure two-party computation (2PC) through this 15-minute conference presentation from USENIX Security '25. Discover how researchers from Tsinghua University and Carnegie Mellon University address the limitations of existing confidential lookup table evaluation methods, which suffer from high asymptotic complexity, practical inefficiency, and sometimes lack proper confidentiality guarantees. Explore the technical details of how FABLE enables a server to hold a confidential lookup table and evaluate it over inputs that are secret-shared between a client and server, producing secret-shared outputs while maintaining privacy. Understand the system's design optimizations for processing multiple inputs with the same lookup table simultaneously, making it particularly valuable for privacy-preserving machine learning inference and data analytics applications. Examine the impressive performance improvements FABLE achieves, delivering 28-101× speedup in LAN environments and 50-393× speedup in WAN environments compared to state-of-the-art confidential lookup table evaluation methods, making secure computation more practical for real-world applications requiring batch processing of sensitive data.