Mohamed Maouche

Dr. Mohamed Maouche

Post-doc at Inria in Privatics Team

PhD from INSA-Lyon with LIRIS Lab

Contact Me Download CV

About Me

I'm Mohamed Maouchec a Post-doc at Inria (Privatics Team), I am currently working on privacy preserving federated learning in DSVD Chaire in partnership with Renault Group. Previously, I was working on private machine learning for speech processing at Inria in the Magnet Team (MAchine learninG in information NETworks). I have receivied a PhD in computer science from INSA-Lyon France in 2019 (working on location privacy).

My main interest is to build machine learning systems that manages a good trade-off between privacy and utility. I explore anonymization techniques and re-identification threats on different data types and application structures.

Professional Experience

Inria

Post-doc - Inria
2021 - Present

Working on privacy preserving federated learning along side Sonia Ben Mokhtar, Antoine Boutet and Jérémie Decouchant. in the DSVD Chaire in partnership with Renault Group. Focusing on health data in the context of car fleets.

Univ-Lille

Teacher - Université de Lille
2021

Teaching Dimension Reduction for the 1st year students of the machine learning master.

Inria

Post-doc - Inria
2019 - 2021

Working on private machine learning for speech processing along side Aurélien Bellet, Marc Tommasi and Emmanuel Vincent.

INSA-Lyon

Teacher - INSA-Lyon
2016 - 2019

Teaches computer science in the computer science department of INSA-Lyon (Dept. IF) and in the first cycle department (Dept PC).

INSA-Lyon

PhD Student - INSA-Lyon with LIRIS Lab
2016 - 2019

In the fields of Data Science, Security and Privacy. Working on Location Privacy and more precisely on re-identification attacks and obfuscation techniques.

INSA-Lyon

Research Intern - Université de Technologie de Compiègne - UTC with Heudiasyc Lab
January 2016 - June 2016

In the field of Optimization in Operations research. Working on the Vehicle Routing Problem (VRP) and the Robust VRP with Time windows constraints.

CDTA

Intern - Centre de Développement des Technologies Avancées of Algiers - CDTA
June 2014 - August 2014

Study of a formal method designed by a research team in the University of Queensland (Australia) which purpose was to transform a BPMN model into a Petri network.

Publications

Enhancing Speech Privacy with Slicing.
[Preprint, submission to Interspeech'22 incoming]
M. Maouche, B. Srivastava, N. Vauquier, A. Bellet, M. Tommasi, E. Vincent.

Privacy and utility of x-vector based speaker anonymization.
[Preprint, submitted to Transactions on Audio, Speech and Language Processing]
B. Srivastava, M. Maouche, Md. Sahidullah, E. Vincent, A. Bellet, M. Tommasi, N. Tomashenko, X. Wang, E. Vincent, J. Yamagishi.

Differentially Private Speaker Anonymization.
[Preprint Soon, submitted to USENIX Security'22]
A. Shamsabadi, B. Srivastava, A. Bellet, N. Vauquier, E. Vincent, M. Maouche, M. Tommasi, N. Papernot.

The VoicePrivacy 2020 Challenge: Results and findings.
[Accepted in Computer Speech and Language]
N. Tomashenko, X. Wang, E. Vincent, J. Patino, B. Srivastava, PG. Noé, A. Nautsch, N. Evans, J. Yamagishi, B. O'Brien, A. Chanclu, JF. Bonastre, M. Todisco, M. Maouche.

A comparative study of speech anonymization metrics.
INTERSPEECH 2020
M. Maouche, B. Srivastava, N. Vauquier, A. Bellet, M. Tommasi, E. Vincent.

Design Choices for X-vector Based Speaker Anonymization.
INTERSPEECH 2020
B. Srivastava, N. Tomashenko, X. Wang, E. Vincent, J. Yamagishi, M. Maouche, A. Bellet, M. Tommasi.

ACCIO: How to Make Location Privacy Experimentation Open and Easy.
ICDCS 2018
V. Primault, M. Maouche, A. Boutet, S. Ben Mokhtar, S. Bouchenak, L. Brunie.

Developped Software

Anonymization Metrics
Integrated to Voice Privacy Challenge 2020

This toolkit encapsulates multiple python implementations of anonymization metrics from the state of the art.

ILL-Attack

A re-identification attack based on learning from multiple past behaviours to re-identify short mobility traces.

HMC

Location Privacy Protection Mechanism HMC (Heat Map Confusion). To protect mobility trace against re-identification attack using heat map profile transformation while maintaining utility.

SFERA

Experiment on re-identification attack on mobility data (AP-Attack, POI-Attack...).

Technical Skills

Advanced Programming Languages

Python (scikit-learn, scipy), Java, Scala, C/C++

Web Languages

HTML, JavaScript, XML, XSL, XPath, XQuery...

System, Network and Database Administration

Linux, MySQL, MongoDB, IOS (cisco)