Artificial Intelligence-based Adaptative Security for Sovreign Clouds – RÉSILIENCE

The RÉSILIENCE LabCom project aims to consolidate a long-standing collaboration between Oodrive, the French leader in sensitive data management, and the L3i laboratory of La Rochelle University. This association, active for more than eight years, has already led to the defense of a doctoral thesis in 2021 and the launch of a new thesis as well as a bilateral collaboration within the framework of France Relance in 2022. Together, the two partners seek to strengthen the security of the Sovereign Cloud and the Trusted Cloud, essential for companies handling sensitive personal and institutional data. This project draws on recent advances in artificial intelligence, including large-scale language models (LLM), privacy-respecting federated learning as well as transfer learning which is essential for addressing not already seen situations. Historically, research on trusted cloud infrastructures has been isolated on three main axes — infrastructure, application services, and digital content — without building a holistic vision. In this LabCom project, we propose an integrated approach where AI will process large volumes of raw data of various types and diverse structuring levels. This will concern on the one hand the protection of sensitive documents. This will be achieved through the development of intelligent mechanisms via LLMs to detect sensitive data in various sectors (e.g. banking, education) and in accordance with regulations, such as the GDPR, in order to prevent data leaks or losses. On the other hand, we will also be interested in data from another level, logs. The objective here will be twofold. Firstly, the idea will be to be able to model the standard behaviors of users in order to allow them to access the resources to which they are entitled according to their context and the degree of sensitivity of the content accessed. Secondly, it will be a question of analyzing these logs in order to detect cyber threat patterns and then to be able to collaborate securely with other cloud players and exchange threat intelligence (CTI). The use of CTI and AI models on these two aspects will not only detect threats, but also proactively improve security by integrating them into existing systems. It is thus, by addressing the three above-mentioned objectives, that we will build the holistic vision targeted by the RESILIENCE LabCom project.