Postdoc opportunity: Federated Learning for Cloud Security

[François-Xavier Dupé]

Hello,

We are looking for a young researcher interested in federated learning issues within a security framework in networks of Cloud type.
In this context, we are looking to design machine learning tools that comply with the GDPR and therefore offer, at the very least,

1. a guarantee of respect for and protection of sensitive data (minimization of data collected and retained, anonymity...) ;
2. explicability of decisions;
3. a guarantee of the right to be forgotten.

In this context, federated learning has several advantages. For example, it does not require the direct sharing
of data, which could be sensitive. Sending data can also be very costly, so we’ll just send the minimum amount
of information necessary. However, securing Cloud-type services, or protecting against denial of service,
requires access to sensitive data (e.g. IP addresses, identifiers) necessary for learning and explicability. A first
research objective will be to find a framework capable of using only useful information (in the GDPR sense).

The theoretical framework of federated learning is fairly generic and can be used for a variety of learning
methods. Nevertheless, it is sensitive to various types of attack that can render the model biased or even unusable.
For example, there are attacks where a malicious site transmits bad information while retaining
good information. A second objective will be to strengthen the model to be robust to adversarial (or Byzantine)
attacks, and to add a network of trust feature.

Location : Laboratoire d’Informatique et des Systèmes, Marseille, FRANCE
Supervisors : François-Xavier Dupé (francois-x...@lis-lab.fr),
                        Emmanuel Godard (emmanue...@lis-lab.fr).
Duration of the post-doc : 1 year, possibility to extend to 2 years to discuss
Deadline for application : July 31st, 2025

For more information, please see the attached file (both in French and English).

Best regards,
François-Xavier Dupé