Robust Federated Learning

Informations générales

Entité de rattachement

Le CEA est un acteur majeur de la recherche, au service des citoyens, de l’économie et de l’Etat.

Il apporte des solutions concrètes à leurs besoins dans quatre domaines principaux : transition énergétique, transition numérique, technologies pour la médecine du futur, défense et sécurité sur un socle de recherche fondamentale. Le CEA s’engage depuis plus de 75 ans au service de la souveraineté scientifique, technologique et industrielle de la France et de l’Europe pour un présent et un avenir mieux maîtrisés et plus sûrs.

Implanté au cœur des territoires équipés de très grandes infrastructures de recherche, le CEA dispose d’un large éventail de partenaires académiques et industriels en France, en Europe et à l’international.

Les 20 000 collaboratrices et collaborateurs du CEA partagent trois valeurs fondamentales :

• La conscience des responsabilités

• La coopération

• La curiosité

  

Référence

2024-33988  

Description du poste

Domaine

Systèmes d’information

Contrat

Stage

Intitulé de l’offre

Robust Federated Learning – Internship H/F

Sujet de stage

This internship focuses on the exploration of Multi-Party Computation (MPC) in a Federated Learning setting, particularly under adversarial conditions. The primary aim is to evaluate how well MPC can maintain local data confidentiality while ensuring effective learning processes in an adversarial environment. This study follows prior work that combined MPC and Federadet learing using clustering techniques to improve performance. The intern will focus on studying the consequences of clustering on system resilience and fault management.

Durée du contrat (en mois)

6

Description de l’offre

Context: Machine learning plays a central role in many applications, and the increasing adoption of decentralized solutions, combined with dependability requirements, necessitates that learning tasks be carried out in a decentralized manner. In such settings, nodes in the system can assume multiple roles: performing learning on their local data while also aggregating the computational results of other nodes.
In this context, we aim to achieve confidentiality, ensuring that private local data is not leaked while providing a practical solution.

 Objective: The goal of this internship is to explore Multi-Party Computation (MPC) in a Federated Learning context, in an adversarial environment, to assess the feasibility and effectiveness of these MPC approaches in ensuring confidentiality, while keeping the learning task effective.
Our team has previously conducted an exploratory study on MPC in the Federated Learning context, in a weak adversarial model (Honest But Curious). The initial global system is composed of n processes and up to fraction of them can be faulty. To improve performance, we proposed a solution where the n processes are organized in clusters, which raises the question of what is a tolerated distribution of faulty processes among the clusters.
The successful candidate will join the Laboratory for Trustworthy, Smart, and Self- Organizing Information Systems (LICIA) at CEA LIST, working in a multicultural, multidisciplinary environment with opportunities to collaborate with external researchers.

 Methodology: The intern will be responsible for the following tasks:
1.     Become familiar with attacks in Federated learning
2.     Conduct a state-of-the-art review of cluster based solutions in the presence of faulty processes.
3.     Definition and analyze different approaches for cluster distribution.
4.     Become familiar with the MPC solution developed in the laboratory.
5.     Implement the diffrent cluster distribution approaches.
6.     Evaluate and compare the performance of the different configurations.

 Requirements:
·        Background in computer science or a related field, with a focus on privacy-preserving technologies, or machine learning.
·        Strong programming skills in languages commonly used for cryptographic or machine learning tasks (e.g., Python, C++).
·        Experience with distributed systems, federated learning, or byzantine faults is a plus.
·        Ability to work independently and collaborate in a research-driven environment.
·        Comfortable working in English, essential for documentation purposes.

Required Specialization: Computer Science
Resources (experiments, analysis methods, others…): destributed systems, programming languages, machine learning, data privacy, fault tolerance.
Desired Level: Master degree (Bac +4/5) – Master 2 Internship
Duration: 6 months
Defense clearance level required (AS minimum): AS
Desired Education: Engineering/Master&apo

Moyens / Méthodes / Logiciels

Distributed systems, programming languages, machine learning, data privacy

Localisation du poste

Site

Saclay

Localisation du poste

France, Ile-de-France, Essonne (91)

Ville

Palaiseau

Critères candidat

Langues

Anglais (Courant)

Diplôme préparé

Bac+5 – Diplôme École d’ingénieurs

Formation recommandée

Informatique

Possibilité de poursuite en thèse

Oui

Demandeur

Disponibilité du poste

10/02/2025