On 20th of June 2023, the Final Review of PRIVILEGE (PRIVacy and Homomorpphic Encryption for Artificial IntElliGencE) Project, which has been funded from PADR 2019 Framework Program of the European Union, was successfully held at Thales Research & Technology premises in Paris, France, with the presence of the Project Officer of the European Commission.
The project’s main objective is originated from the fact that data sharing between the members of a coalition (e.g. NATO) is unquestionably beneficial for training AI algorithms. However, unrestricted sharing of sensitive military data between military partners is currently unthinkable, even on small scale exercises.
The project advanced the state-of-the-art of AI enabled defence systems, with respect to the data security and privacy preserving AI collaborative approaches. PRIVILEGE provides a framework for sharing of AI models, while at the same time enhances model training on encrypted data across the military partners without disclosing information of any kind. The technological bricks rely on the use of Homomorphic Encryption (HE), Verifiable Computing (VC), Federated Learning (FL) and Private Aggregation of Teacher Ensembles technologies (PATE).
PRIVILEGE solutions were validated on three defence originated use-cases, namely, waveform recognition, network cybersecurity and image classification for ISR missions based on accuracy, integrity and performance metrics.
In the context of PRIVILEGE, INTRACOM DEFENSE (IDE) was responsible for the execution of WP2, Use Cases and Architecture, as well as the adaptation of the PRIVILEGE framework to the image classification for ISR missions use case.
It was a great pleasure and honour for IDE to be part of this journey, along with the PRIVILEGE partners THALES SIX GTS (THA/FR-Coordinator), Comissariat à l’énargie atomique et aux énergies alternatives (CEA/FR) and CESNET, association of legal entities (CESNET/CZ).