Missions

The handling of sensitive data is usually performed by a trusted machine that must first decrypt the stored data before processing it. This preliminary step creates a break in the secure storage-computation continuum. Furthermore, although data processing can be carried out in a hardware enclave, many operations, such as those related to machine learning, require the use of GPUs. This is the case, for example, for actors who wish to collaborate (e.g., through the training of a federated learning model) but do not trust each other to use a private and distributed learning tool on top of enclave-secured storage. The objective of this post-doctoral project is to develop privacy-preserving federated learning algorithms in the edge-cloud continuum. One promising approach is to leverage recent secure technologies such as NVIDIA confidential computing (GPUs combined with secure enclaves) to achieve this continuum.

Activités

• Conduct a state-of-the-art review of privacy-preserving distributed learning techniques in the edge-cloud continuum and identify the limitations of existing solutions.
• Propose a novel privacy-preserving federated learning solution within the edge-cloud continuum.
• Implement the proposed solution and evaluate its performance and resilience to attacks.
• Publish a paper presenting the obtained results

Compétences

A PhD in AI with a preference to distributed AI is required.

Contexte de travail

The LIRIS (Laboratoire d'InfoRmatique en Image et Systèmes d'information) is a joint research unit (UMR 5205) of the CNRS, INSA Lyon, Claude Bernard Lyon 1 University, Lumière Lyon 2 University, and Centrale Lyon. It has 330 members.

The postdoctoral researcher will join the DRIM team at LIRIS.

Le poste se situe dans un secteur relevant de la protection du potentiel scientifique et technique (PPST), et nécessite donc, conformément à la réglementation, que votre arrivée soit autorisée par l'autorité compétente du MESR.

Contraintes et risques

None