Dario Fiore, research professor at IMDEA Software, has been awarded an ERC Proof of Concept (PoC), a grant from the European Research Council aimed at translating high-level scientific results into applications with social and industrial impact. This type of ERC grant is exclusively available to researchers who have previously led ERC projects. Its purpose is clear: to explore how scientific advances can be transformed into useful real-world solutions.
The project awarded to Dario Fiore has a budget of 150.000€ and a duration of 18 months, and it 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, titled “VERIFHE: An Open-Source Library for Verifiable Fully Homomorphic Encryption for Trustworthy Machine Learning”, addresses a problem that is increasingly present in our daily lives. Many organizations now rely on external Machine Learning as a Service (MLaaS) providers to analyze data and generate predictions. This requires sending information, often sensitive, to third parties.
Cryptographic techniques exist to protect the confidentiality of such data. Among them, homomorphic encryption stands out, allowing computations to be performed directly on encrypted data. This means the service provider never has access to the original information. However, this approach leaves a key question unanswered: How can a user be sure that the returned result is correct? The project led by Dario Fiore during his ERC, called PICOCRYPT, addressed this question.
One of the main contributions of the PICOCRYPT project is the incorporation of cryptographic verification mechanisms into homomorphic encryption. In practice, this not only allows the privacy of the data to be preserved, but also ensures that computations have been carried out correctly, even if the provider makes mistakes or acts maliciously.
VERIFHE thus builds on these recent advances developed within this research, where a new verifiable homomorphic encryption protocol was proposed. The ERC Proof of Concept will now enable the transformation of this theoretical result into a practical and reusable tool.
The project has two concrete objectives: on one hand, to develop the first open-source library implementing verifiable homomorphic encryption; on the other hand, to build a prototype of a third-party machine learning service that demonstrates the feasibility of performing private and verifiable inference. These outcomes may be particularly relevant in sectors where data privacy and result accuracy are critical, such as in many digital services handling 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.
Funded by the European Union (GA: 101287502). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.
