Skip to main content
Go to the home page of the European Commission (opens in new window)
English en
CORDIS - EU research results
CORDIS

An Open-Source Library for Verifiable Fully Homomorphic Encryption for Trustworthy Machine Learning

Project description

Protecting sensitive data from malicious tampering

Sending sensitive data to third-party providers for cloud-based machine learning as a service – where companies outsource their AI data processing – exposes users to severe privacy breaches and regulatory penalties. Fully homomorphic encryption solves this privacy issue by processing data while remaining encrypted but lacks a mechanism to verify that the calculated results are actually correct. This vulnerability leaves users highly exposed to hidden errors or malicious data manipulation. The ERC-funded VERIFHE project will develop an open-source library that will introduce verifiable homomorphic encryption technology. Building upon recent cryptographic breakthroughs, researchers will create a proof-of-concept system that combines privacy protection with a verification mechanism. The resulting tool will enable users and industries to confidently adopt secure, trustworthy machine learning services.

Objective

Machine Learning as a Service (MLaaS) is emerging as a cornerstone of the modern digital infrastructure, but its widespread adoption is hampered by critical privacy and integrity concerns. Users must transmit sensitive data to third-party providers, risking privacy breaches and regulatory violations. While Fully Homomorphic Encryption (FHE) offers a powerful solution for privacy by enabling computations on encrypted inputs, it provides no way to verify the correctness of the results. This leaves users vulnerable to errors or malicious manipulation with potentially severe consequences.

The VERIFHE project addresses this critical gap in privacy-preserving MLaaS by providing integrity guarantees through verifiable FHE, a cryptographic primitive that, in addition to enabling computations on encrypted inputs, also provides a mechanism to check their correctness. VERIFHE builds upon a recent breakthrough from the ERC-funded PICOCRYPT project—a novel protocol for verifiable FHE—and aims to transform it into a practical, opensource tool. The project’s primary outcomes will be the first-ever open-source software library for verifiable FHE, and a proof-of-concept MLaaS system that, using the library, offers verifiable and privacy-preserving machine learning inference.

By delivering the first open-source library for verifiable FHE and demonstrating its feasibility in MLaaS, VERIFHE will accelerate adoption of this technology and pave the way for trustworthy machine learning in industrial sectors where privacy and integrity are paramount.

Fields of science (EuroSciVoc)

Keywords

Programme(s)

Topic(s)

Funding Scheme

HORIZON-ERC-POC - HORIZON ERC Proof of Concept Grants

See all projects funded under this funding scheme

Call for proposal

(opens in new window) ERC-2025-POC

See all projects funded under this call

Host institution

FUNDACION IMDEA SOFTWARE
Net EU contribution
€ 150 000,00
Address
CAMPUS DE MONTEGANCEDO SN
28223 Pozuelo De Alarcon
Spain
Region
Comunidad de Madrid Comunidad de Madrid Madrid
Activity type
Research Organisations
Links
Total cost
No data

Beneficiaries (1)