ComputeFHE: A Privacy-Preserving General-Purpose Computation Library

ComputeFHE is an open-source C++ library designed to lower the barrier to entry for developing privacy-preserving applications using Fully Homomorphic Encryption (FHE). By building on the TFHE cryptosystem and the OpenFHE backend, it allows developers to write code using familiar imperative programming paradigms—such as standard integer and fixed-point data types, conditional branching, and vector operations—while the library handles the underlying cryptographic complexity.

The library provides two primary modes of operation: a standard ALU implementation based on conventional two-input logic gates and an optimized ALU implementation that utilizes specialized three-input logic gates. These optimized primitives are designed to reduce the frequency of bootstrapping operations, which are the primary computational bottleneck in FHE. Additionally, ComputeFHE includes a simulation mode that allows developers to test, debug, and analyze the circuit complexity and bootstrapping costs of their algorithms without performing the actual, resource-intensive cryptographic computations.

Experimental results indicate that the optimized ALU architecture significantly outperforms the standard approach across a variety of operations. For example, the library achieves performance gains of up to 3.9x for specific comparison operations and nearly 2x for sorting algorithms like the Batcher odd-even mergesort. By automating the management of encrypted state and providing intuitive abstractions, ComputeFHE enables developers to implement complex algorithms—such as those requiring oblivious array access—without requiring deep expertise in the underlying cryptographic mechanisms.

FHE is a powerful tool for data privacy, but its adoption has been historically hindered by high development complexity and extreme computational overhead. ComputeFHE addresses these challenges by providing a practical framework that balances ease of use with performance-oriented optimizations. Its ability to estimate costs via simulation and its support for arbitrary bit widths make it a valuable tool for researchers and developers aiming to integrate privacy-preserving computations into real-world applications.