Benchmarking homomorphic encryption libraries with a focus on multithreading

Abstract

Homomorphic encryption (HE) enables computations on encrypted data without requiring decryption, offering strong privacy guarantees for sensitive applications such as healthcare, finance, and machine learning. Among the various HE schemes, Cheon-Kim-Kim-Song (CKKS) stands out as one of the most widely used and well-suited for performing approximate arithmetic on real or complex numbers. This study presents a comparative benchmark of three state-of-the-art HE libraries supporting the CKKS scheme, i.e., Microsoft SEAL, OpenFHE, and Lattigo, with a particular emphasis, for the first time, on multi-threaded performance. In addition to fundamental operations such as encryption, key generation, additions, and rotations, crucial operations such as inner product and matrixvector multiplication were evaluated under various parameter settings and thread configurations. Results reveal distinct tradeoffs among the libraries in terms of latency and multithreading capabilities, offering valuable insights for selecting the most appropriate HE library for real-world, performance-sensitive, privacy-preserving applications. Our implementation is available at https://github.com/ErenOzilgili/FHE_libs. © 2025 IEEE.