A comparative technical benchmark of leading open-source homomorphic encryption frameworks applied to high-throughput identity verification.
Homomorphic encryption has advanced from theoretical curiosity to practical production capability. We examine the architecture, SIMD batching capabilities, and GPU acceleration features of OpenFHE, Microsoft SEAL, and Zama Concrete.
Detailed empirical benchmarks measuring execution time, memory footprints, and ciphertext expansion ratios when processing 512-dimensional floating-point vectors on modern multi-core x86 and ARM servers.
Demonstrating how GPU offloading (NVIDIA CUDA) accelerates polynomial NTT operations in homomorphic evaluation by over 40x, enabling cloud servers to process over 500 biometric verifications per second per server node.