The ERATOSTHENES project, in which Eviden led the “Decentralized Identity Management” work package and the “Federated Threat Analysis Models for Continuous Risk Assessment” task, ended successfully on March 31, 2025 after three and a half years of work involving colleagues from the Identity Management and Privacy Team and the Trustworthy AI Team. 

To close it, a final review was held on 20th of May in Brussels, Belgium. The consortium presented there the outcomes of the project, including the results from the final period of ERATOSTHENES. 

The project has completely achieved the expected results but implementing privacy-preserving technologies such as distributed attribute-based credentials (dP-ABC) introduced complexity and integration difficulties, which the team overcame by developing modular and integrable solutions. The integration of new Trusted Execution Environments (TEEs) and Physical Unclonable Functions (PUFs) demanded rigorous testing, but ultimately established a robust, hardware-rooted trust layer.

The main result of the Identity Management and Privacy Team is related to the Ledger uSelf component. It provides a set of tools to support the adoption of a Decentralized Identity Management solution within the Eratosthenes project, following a Self-Sovereign Identity (SSI) paradigm. The core mission was to reimagine identity for the Internet of Things (IoT) by moving away from centralized identity providers in favor of a decentralized, self-sovereign identity approach for both devices and users.

The Ledger uSelf component is based on the implementation of W3C Decentralized Identifiers (DIDs) and W3C Verifiable Credentials (VCs) standards. It is compliant with GDPR and aligned with the European Self-Sovereign Identity strategy.

The solution was designed to maintain privacy, control, and resilience by integrating advanced cryptographic techniques such as dP-ABC, and by supporting secure identity data storage through the Trusted Execution Environment of the IoT devices. This enables decentralized trust frameworks and establishes mechanisms for secure identity recovery. To ensure both privacy and interoperability, the solution leverages Verifiable Credentials and Decentralized Identifiers, promoting transaction anonymity and adherence to the principle of minimal data disclosure.

The main result of the Trustworthy AI team in ERATOSTHENES is related to the FedLPy component which is a Python package that implements a Federated Learning framework for the continuous detection of malware traffic in IoT networks. The component is comprised of two submodules: the FL system and the Continuous Assessment block, both with an implementation of a client and a server agent, to fit different participant roles in the federation.

FedLPy enables training AI models with raw data from IoT networks and exploiting this knowledge to detect Denial-of-service (DDoS) attacks on a continual basis. Because the system aims to run on IoT devices, FedLPy is very lightweight and optimized to minimize the bandwidth consumption during training.
 
Eviden’s contribution has been key to the project because:

• The uSelf Ledger component plays a crucial role in the project, as it focuses on digital identity, and this component provides decentralized digital identity to IoT devices. This contribution has been essential, as it delivers a key functionality to the overall system. In addition, it serves as an integration point for components from other work packages, as well as from within its own work package, ensuring cohesion and interoperability across the project.
• FedLPy is a key component of the Intrusion Detection System (IDS) of the ERATOSTHENES architecture, which adds an extra layer of security by leveraging FL techniques to train AI models capable of detecting threats in real time. These models are trained with privacy preservation techniques and using raw packets from IoT networks, which constitutes a novel approach with respect to state-of-the-art anomaly detection systems, enhancing the innovation aspect of the project.

Eviden has a strong portfolio in cryptography and Identity and Access Management (IAM) solutions, serving various sectors. In 2024, it launched Evidian Orbion, a next-generation Identity-as-a-Service (IDaaS) platform that combines Identity Governance, Access Management, and Privileged Access Management. This solution supports identity proofing and plans to integrate decentralized identities (SSI) to meet growing demands for Self-Sovereign Identity and Verifiable Credentials. The company will be able to incorporate the results of the ERATOSTHENES project into its offerings, enhancing access control and governance within its IAM and Trusted Identity solutions for IoT.

On the other hand, FedLPy component represents the first steps of the Trustworthy AI team in the Federated Learning research line, so the technical developments behind FedLPy set the bases for the Federated Learning framework that is currently being developed by this unit.