Dario Fiore, research professor at IMDEA Software, has been awarded an ERC Proof of Concept (PoC) grant by the European Research Council. This funding scheme is designed to help translate high-level scientific results into applications with social and industrial impact and is exclusively available to researchers who have previously led ERC-funded projects.

The awarded project has a budget of €150,000 and a duration of 18 months, and will focus on one of the major challenges in digital technology today: ensuring both privacy and reliability in cloud-based machine learning services.

The project, entitled “VERIFHE: An Open-Source Library for Verifiable Fully Homomorphic Encryption for Trustworthy Machine Learning”, addresses a problem that is increasingly present in everyday digital services. Many organizations rely on external Machine Learning as a Service (MLaaS) providers to analyze data and generate predictions, a process that requires sending often sensitive information to third parties.

While cryptographic techniques exist to protect data confidentiality, homomorphic encryption stands out for enabling computations to be performed directly on encrypted data, preventing service providers from accessing the original information. However, this approach raises a fundamental question: how can users be sure that the results they receive are correct?

This challenge was addressed in PICOCRYPT, an ERC-funded project led by Dario Fiore, which introduced cryptographic verification mechanisms into homomorphic encryption. These mechanisms make it possible not only to preserve data privacy, but also to verify that computations have been carried out correctly, even in the presence of errors or malicious behavior by the service provider.

The VERIFHE project builds on these advances, which include the proposal of a new protocol for verifiable homomorphic encryption. The ERC Proof of Concept grant will now enable the transformation of this theoretical result into a practical and reusable tool.

VERIFHE has two concrete objectives: first, to develop the first open-source library implementing verifiable homomorphic encryption; and second, to build a prototype of a third-party machine learning service demonstrating the feasibility of private and verifiable inference.

These outcomes may be particularly relevant in sectors where data privacy and result accuracy are critical, such as digital services that handle sensitive information.

With this ERC Proof of Concept, Dario Fiore’s team aims to advance the development of technologies that contribute to a safer and more reliable digital landscape.