Nemesis: Noise-randomized Encryption with Modular Efficiency and Secure Integration in Machine Learning Systems

TL;DR

Nemesis is a novel framework that enhances the efficiency of Fully Homomorphic Encryption in machine learning systems, enabling faster computations while maintaining security and accuracy, thus facilitating broader adoption of privacy-preserving ML.

Contribution

Nemesis introduces advanced caching techniques and mathematical tools to accelerate FHE operations, supporting general plaintext structures and improving upon prior caching-based methods.

Findings

Significantly reduces FHE computational overhead.

Maintains accuracy and security in ML tasks.

Effective on datasets like MNIST, FashionMNIST, CIFAR-10.

Abstract

Machine learning (ML) systems that guarantee security and privacy often rely on Fully Homomorphic Encryption (FHE) as a cornerstone technique, enabling computations on encrypted data without exposing sensitive information. However, a critical limitation of FHE is its computational inefficiency, making it impractical for large-scale applications. In this work, we propose \textit{Nemesis}, a framework that accelerates FHE-based systems without compromising accuracy or security. The design of Nemesis is inspired by Rache (SIGMOD'23), which introduced a caching mechanism for encrypted integers and scalars. Nemesis extends this idea with more advanced caching techniques and mathematical tools, enabling efficient operations over multi-slot FHE schemes and overcoming Rache's limitations to support general plaintext structures. We formally prove the security of Nemesis under standard cryptographic assumptions and evaluate its performance extensively on widely used datasets, including MNIST, FashionMNIST, and CIFAR-10. Experimental results show that Nemesis significantly reduces the computational overhead of FHE-based ML systems, paving the way for broader adoption of privacy-preserving technologies.