Detailed information about the course

[ Back ]
Title

Secure and Efficient Learning: approaches, techniques and threats

Dates

1-2 Decembre 2021

Lang EN Workshop language is English
Responsable de l'activité

Valerio SCHIAVONI

Organizer(s)

Dr Valerio Schiavoni, UniNE

Pr Christian Cachin, UniBE

Speakers

Dr Giorgia Marson, NEC Labs, Germany. Bio: Giorgia Marson is currently a senior researcher at NEC Labs Europe. She received her M.Sc. in Mathematics from Sapienza University of Rome, Italy, in 2011, and her PhD from TU Darmstadt, Germany, in 2016. As a PhD student, she has been working under the supervision of Marc Fischlin, focusing on modeling and analyzing cryptographic protocols. Afterwards, she has worked as postdoc at Ruhr University Bochum, Germany, as research scientist at NEC Labs Europe, and as postdoc at University of Bern, Switzerland. Broadly, her research interests include cryptography, information security, and distributed systems, with current focus on blockchain and machine-learning security.

Dr Sonia Ben Mokhtar, University of Lyon, LIRIS, CNRS, France. Bio: Sonia Ben Mokhtar has been a CNRS researcher at the LIRIS lab since October 2009. Since 2017, she is also the leader of the distributed systems and information retrieval group (DRIM). Before joining CNRS, She was a research associate at University College London (UCL) for two years, working with Licia Capra. She received her PhD in 2007 from University Pierre et Marie Curie (Paris 6), under the supervision of Valérie Issarny and Nikolaos Georgantas in the former INRIA ARLES project-team (currently MiMove).

Dr Mathias Humbert, University of Lausanne, Switzerland. Bio: Mathias Humbert is a scientific project manager at the Cyber-Defence Campus (Switzerland) and will join the University of Lausanne as an associate professor in November 2021. Prior to this, he was a senior data scientist at the Swiss Data Science Center (ETH Zurich, EPFL) and a post-doctoral researcher at the Center for IT-Security, Privacy, and Accountability (CISPA) in Saarbrücken, Germany. He completed his Ph.D. thesis on privacy protection in early 2015 in the School of Computer and Communication Sciences at EPFL, after M.Sc. (2009) and B.Sc. (2007) studies at EPFL and UC Berkeley. He is a recipient of the NDSS 2019 distinguished paper award.

Pr Alain Tchana, ENS-Lyon, France . Bio: I studied in Cameroon (Africa) until 2008. I am graduated from University of Yaoundé I. Then I received my PhD in computer science in 2011, at the IRIT laboratory, Institut National Polytechnique de Toulouse, France. Since September 2019 I'm a Full Professor at Ecole Normale Supérieure de Lyon, France. My main research interests are in Virtualization, Operating Systems, and Cloud Computing in general. Currently, I am mainly interested in datacenter disaggregation, storage, virtualization of hardware features for virtualization and, in securing applications. Simply, I'm the Systems guy!

Dr Vlad Nitu, CNRS, INSA-Lyon, France. Bio: Vlad Nitu is a junior researcher (chargé de recherche) at The French National Centre for Scientific Research, working on the energy-efficiency and the robustness of Federated Learning. Previously, he was a Postdoctoral Researcher at EPFL - Lausanne, working in the team of Prof. Rachid Guerraoui. Vlad obtained his PhD from Toulouse University under the supervision of Prof. Daniel Hagimont and Prof. Alain Tchana on the energy-efficiency of virtualized cloud datacenters.

Peterson Yuhala, UniNE, Switzerland. Bio :Peterson Yuhala received his Computer Engineering diploma from the National Advanced School of Engineering, Cameroon, in 2018. He is presently a third year PhD student at the University of Neuchatel, Switzerland, under the supervision of Prof. Pascal Felber, Dr. Valerio Schiavoni and Prof. Alain Tchana. His research interests are in efficient privacy preserving computation with trusted execution environments (TEEs) and persistent memory (PM).

Description

This CUSO seminar will revolve around security and efficiency concerns in the context of machine learnin, including novel scenarios such as federated learning. Our speakers will engage with students presenting recent research works on a set of highly correlated domains, as well as offering interactive sessions.

We prospect the following talks from our invited speakers:

Dr Mathias Humbert (UNIL, Switzerland): Machine Learning for Privacy, Privacy for Machine Learning.

Dr Sonia Ben Mokhtar (INSA/CNRS/U. Lyon) will offer a lecture on privacy threats for federated learning schemes, as well as surveying some of the possible attacks that such decentralized systems can suffer.

Dr Vlad Nitu (CNRS. France), in the same team as Dr Sonia Ben Mokhtar, will describe the challenges in designing energy-efficient machine learning systems in particular when deployed at the edge.

Dr Giorgia Azzurra Marson (NEC Labs Europe, Germany) will give an overview on "adversarial ML", showing how attackers can fool ML classifiers in the context of evasion and poisoning scenarios. In the first lecture, we will focus on selected attack strategies under different threat models. In the second lecture, we will explore defensive strategies and discuss challenges and open problems in the field.

 Pr Alain Tchana (U. Lyon, France) will present his recent work on Plinius, a system that leverages TEEs to shield models from malicious or compromised systems that can be used over untrusted cloud providers.

In addition, we intend to offer a live and interactive session on how to use this system, given by Peterson Yuhala, PhD student at the University of Neuchatel.

Program

Lectures will be in person-only. All lectures will happen in the UniMAIL building, Room E-301-304.

Final program below.

Wednesday 01.12 UNIMAIL, Room E-301-304.     
10:00 Intro/welcome day-1 15 min.  
10:15 Lecture 1.1: Machine Learning for Privacy, Privacy for Machine Learning (1/2) 45min. Pr Mathias Humbert (UNIL, Switzerland)
11:00 Break 15 min  
11:15 Lecture 1.2: Machine Learning for Privacy, Privacy for Machine Learning (2/2) 45min. Pr Mathias Humbert (UNIL, Switzerland)
12:00 Lunch (self or catering) 1h30min  
13:30 Lecture 1.3: Privacy-Preserving Collaborative Learning 45 min. Dr Sonia Ben Mokhtar (INSA/CNRS/U.Lyon, France)
14:15 End of day 1    
       
       
Thursday 02.12 UNIMAIL, Room E-301-304.    
08:55 Intro/welcome day-2    
09:00 Lecture 2.1: Energy-Efficient Federated Learning 45 min. Dr Vlad Nitu (CNRS/INSA- Lyon, France)
09:45 Break 15min  
10:00 Lecture 2.2: ML classification in adversarial setting: how to fool ML classifiers 45 min Dr Giorgia Azzurra Marson (NEC Labs Europe, Germany)
10:45 Break 15 min  
11:00 Lecture 2.3: Robust ML classification: defensive techniques and challenges 45 min Dr Giorgia Azzurra Marson (NEC Labs Europe, Germany)
12:00 Lunch (self or catering) 1h30min  
13:30 Lecture 2.4: Plinius: Secure and Persistent Machine Learning Model Training 45 min. Pr Alain Tchana (ENS Lyon, France)
14:15 Break 15min  
14:30 Lecture 2.5: Plinius Live Demo 45 min Peterson Yuhala (UniNE, Switzerland)
15:15 Discussion and closing aperitif 15/30min
Location

Neuchatel, UniMail Building (room E301-304)

Map

Map

Information
Places

30

Deadline for registration 30.11.2021
Contact

valerio.schiavoni@unine.ch

short-url short URL

short-url URL onepage