Preparation of a Eurogrant application planned under the Horizon Europe programme

alpha0 sp. z o. o. is implementing a project co-financed by the European Funds entitled “Preparation of a Eurogrant application planned under the Horizon Europe programme”. Project number: FENG.02.12-IP.02-0285/26.

What is the aim of the project?

The aim is to prepare a complete application package and the necessary analyses to support the submission of a high-quality Eurogrant proposal concerning the development of new solutions in robust federated learning for cybersecurity applications.

What will the planned Eurogrant cover?

The planned Eurogrant will focus on developing a concept, methods and a research prototype for robust federated learning intended for systems that detect threats, incidents and anomalies.

The solution is intended to enable multiple participants to train models collaboratively without centrally collecting their data, while limiting the impact of erroneous or deliberately manipulated data and updates on the resulting model.

What is the objective of the planned R&D work?

The objective of the R&D work is to create new methods that improve the resilience of federated learning to disruptions and deliberate abuse. The work will involve developing a threat model, investigating the conditions under which the learning process loses its robustness, developing methods to detect attempted manipulation and designing mechanisms to limit their impact on model quality and stability.

Scope of the planned Eurogrant

In particular, the planned Eurogrant will include:

• developing a threat model for federated learning used in cybersecurity systems;
• identifying attack scenarios, including data poisoning, manipulation of model updates, introduction of backdoors and impersonation of other participants in the learning process;
• developing methods for robust aggregation of model updates submitted by multiple participants;
• developing mechanisms to detect suspicious, erroneous or deliberately manipulated updates;
• developing methods to limit the impact of malicious participants on the learning process and its results;
• developing methods to assess the trustworthiness of data and updates originating from individual participants;
• preparing mechanisms to document the learning process and improve the accountability of the solution;
• developing metrics and procedures to evaluate the effectiveness, quality, stability and robustness of the proposed methods;
• preparing a research prototype of a robust federated learning system;
• conducting laboratory tests covering both normal system operation and scenarios involving deliberate attempts to disrupt learning.

What will the planned Eurogrant deliver?

The planned Eurogrant is expected to deliver a threat model for federated learning in cybersecurity, a set of new methods that increase resilience to erroneous and malicious data, mechanisms to detect and mitigate manipulation attempts, quality and robustness assessment procedures, a research prototype and technical documentation enabling further validation and technological development.

What will this project deliver?

The result of this project will be a complete Eurogrant proposal ready for submission and meeting the formal and substantive requirements of the European Commission.

This will enable the applicant to seek funding for R&D into new robust federated learning methods that could improve the trustworthiness and security of analytical systems used to detect cyber threats in distributed environments.

Project funding