Fellows and Projects
Fellow 1: Théo Nguyen
Speech anonymisation for privacy-preserving emotion recognition
About me: I come from France where I studied music technology at IRCAM/Sorbonne University. After a short experience as a research engineer at Evry University where I worked on behavioural analysis with speech emotion recognition in healthcare interactions, here I am for this doctoral adventure. My journey starts in Aalto University (Finland) then ends in Ruhr-Universität Bochum (Germany), supervised by Prof. Tom Bäckström and Prof. Dr.-Ing. Rainer Martin. My topic is about speech anonymization for privacy-preserving emotion recognition
Institutions: Aalto University (Finland), Ruhr University Bochum (Germany)
Secondment: HEAD acoustics
Supervisors: Tom Bäckström (Aalto University), Rainer Martin (RUB)
Contact: theo.nguyen@aalto.fi
Fellow 2: [To be announced]
Disentangled representations for selective attribute suppression
Institutions: EURECOM (France), Ruhr University Bochum (Germany)
Secondment: Orange
Supervisors: Nicholas Evans (EURECOM), Rainer Martin (RUB), Nicolas Gengembre (Orange)
Fellow 3: Priyanshi Pal
Protection against deepfakes in speech
About me: The goal of my research is towards developing protective measures against audio deepfakes through proactive defenses, such as watermarking and optimized perturbations, that can disrupt voice-cloning outputs. This is driven by the need for protection at the source, against voice cloning, apart from reliance only on post-production detection of synthetic audio.
Institutions: INESC-ID (Portugal), Aalto University (Finland)
Secondment: VoiceMod
Supervisors: Isabel Trancoso (INESC-ID), Lauri Juvela (Aalto University), Jordi Janer (VoiceMod), Pritish Chandna (VoiceMod)
Contact: priyanshi.pal@aalto.fi
Fellow 4: Jiusi Zheng
Transparent Exchange of Speaker Attributes
About me: Jiusi Zheng is a PhD candidate working on transparent speaker attribute exchange, aiming to protect attributes that are hard for people to perceive. His research interests span the fields of speech processing and privacy-preserving machine learning, with an emphasis on developing interpretable methods that balance utility and privacy in speech processing.
Institutions: Radboud University (Netherlands), Aalto University (Finland)
Secondment: VoiceInteraction
Supervisors: Martha Larson (Radboud University), Tom Bäckström (Aalto University), Carlos Mendes (VoiceInteraction)
Contact: jiusi.zheng@ru.nl
Fellow 5: Anastasiia Korenevskaia
Revealing social relationships in conversations
About me: I’m originally from Saint Petersburg, Russia. My background is in mathematics from my bachelor’s studies, during which I explored Gaussian processes. These turned out to be closely related to speech processing, and this area that sparked my curiosity. To deepen my knowledge of speech technologies, I joined the Master’s program in Computer Science and Speech Technology at ITMO University in Russia. At the same time, I started working as a junior researcher at Speech Technology Center as part of a speaker verification team. Both my studies and work gave me hands-on experience with different aspects of speech. Now I’m investigating another side of speech that I find unexpected and captivating: speech privacy. Specifically, I’m looking into speech privacy in multi-speaker conditions.
Institutions: Radboud University (Netherlands), INESC-ID (Portugal)
Secondment: Njita
Supervisors: Martha Larson (Radboud University), Alberto Abad (INESC-ID)
Contact: anastasiia.korenevskaia@ru.nl
Fellow 6: Shilpa Chandra
Robust attack models and tools for the credible evaluation of anonymisation and attribute suppression
About me: Shilpa is a PhD researcher working on voice anonymisation and speech privacy. Her research focuses on understanding whether voice anonymisation methods are actually effective by developing strong attack models to evaluate them. Her research also focuses on explaining and tracing which parts of speech may still remain identifiable after anonymisation.
Her PhD research is being conducted jointly at EURECOM/Sorbonne University and TU Berlin.
Institutions: EURECOM (France), Technische Universität Berlin (Germany)
Secondment: Omilia
Supervisors: Nicholas Evans (EURECOM), Dorothea Kolossa (TUB), Themos Stafylakis (Omilia)
Contact: shilpa.chandra@eurecom.fr
Fellow 7: Eulalie Thiombiano
Privacy impact assessment for comprehensive attacks exploiting audio, speech, and metadata
About me: I am a PhD researcher in the PSST doctoral network, working on privacy impact assessment for speech technologies. My research explores how privacy risks emerge when speech data is combined with additional information such as metadata, spoken content, video, or external side information.
My goal is to move beyond audio-only privacy evaluation and study more realistic attack scenarios, where speakers may be re-identified or sensitive traits may be inferred from multiple sources of information. Through this work, I aim to contribute to more robust threat modelling and privacy evaluation frameworks for responsible speech technologies.
Institutions: INRIA (France), Radboud University (Netherlands)
Secondment: CNIL
Supervisors: Emmanuel Vincent (INRIA), Vincent Colotte (Université de Lorraine / LORIA), Martha Larson (Radboud University), Romain Darous (CNIL), Vincent Toubiana (CNIL)
Contact: eulalie.thiombiano@ru.nl
Fellow 8: Dāvis Šterns
Attacking information bottlenecks – Theoretical metrics and bounds of privacy
About me: I am originally from Latvia and now work as a researcher at the intersection of privacy and speech technologies. My research is driven by the increasing need to protect people from powerful algorithms that can silently extract sensitive information from our digital data. Building on my background in Data Science and Privacy-Enhancing Technologies, I currently study how privacy leakage in anonymised speech can be measured using information-theoretic methods and ideas from Quantitative Information Flow.
Institutions: Aalto University (Finland), INRIA (France)
Secondment: Nokia
Supervisors: Tom Bäckström (Aalto University), Catuscia Palamidessi (INRIA)
Contact: davis.sterns@aalto.fi
Fellow 9: Hoang Pham
Robust privacy-preserving industrial voice interfaces
About me: I am originally from Vietnam and have a background in Artificial Intelligence, developed through both my academic training and working experience as an AI engineer. My research journey began during my Master’s studies at the University of Eastern Finland, where I focused on the security in speech. My Master’s thesis investigated speech anti-spoofing, laying the foundation for my continued interest in trustworthy and secure speech technologies.
Currently, I am a doctoral researcher in the Privacy for Smart Speech Technology (PSST) project, funded by the Marie Skłodowska-Curie Actions. My PhD research is jointly supervised by Prof. Dr.-Ing. Rainer Martin at Ruhr University Bochum, Germany, and Prof. Lauri Juvela at Aalto University, Finland. My work focuses on developing privacy-preserving solutions for speech technologies in challenging real-world conditions involving background noise, multiple speakers, and diverse recording device
Institutions: Ruhr University Bochum (Germany), Aalto University (Finland)
Secondment: ELDA
Supervisors: Rainer Martin (RUB), Lauri Juvela (Aalto University)
Contact: hoang.pham@edu.ruhr-uni-bochum.de
Fellow 10: Benjamin Luks
Detection of speech-affecting diseases in anonymized speech
About me: Ben is researching methods of detecting diseases in anonymized speech. He has backgrounds in speech technology, linguistics and music. This research will be conducted at INESC-ID in Lisbon under the supervision of Dr. Alberto Abad and Dr. Isabel Trancoso, and at Technische Univerität Berlin with Dr.-Ing. Sebastian Möller. He will also be pursuing a secondement at ki:elements under the supervision of Dr. Johannes Tröger.
Institutions: INESC-ID (Portugal), Technische Universität Berlin (Germany)
Secondment: ki:elements
Supervisors: Alberto Abad (INESC-ID), Sebastian Möller (TUB), Johannes Tröger (ki:elements)
Contact: benluks@inesc-id.pt
Fellow 11: Chenyi Lin
Utility of Speech Samples as Privacy-Preserving, Transparent and Reusable Model-Updates for Distributed Learning
About me: I am a PhD researcher in the PSST doctoral network, with a background in linguistics and speech technology. My research focuses on privacy-preserving and transparent distributed learning for speech technologies. In particular, I explore how speech-based representations of model updates can improve the efficiency, interpretability, and privacy of federated learning systems. More broadly, I am interested in developing trustworthy and human-centered speech technologies.
Institutions: Aalto University (Finland), EURECOM (France)
Secondment: ELDA
Supervisors: Tom Bäckström (Aalto University), Nicholas Evans (EURECOM)
Contact: chenyi.lin@aalto.fi
Fellow 12: Victor Ménestrel
Methods for subjective and objective evaluation of privacy
About me: I am a PhD candidate in machine learning working on speech quality evaluation and privacy-preserving speech technologies.
I am part of the PSST (Privacy for Smart Speech Technology) project funded by the Marie Skłodowska-Curie Actions, and I conduct my research at TU Berlin within the MTEC Lab and the Quality & Usability Lab.
I focus on evaluating speech characteristics such as intelligibility, emotional expression, and overall perceptual quality using both objective (Machine/Deep Learning) and subjective (Human perception) methods. Because speech encode biometric information, speech anonymization is necessary to protect privacy. However, simply changing voice timbre is not enough to remove all biometric data, as factors like intonation, rhythm, and accent can still reveal the identity of speakers. At the same time, removing all of this information can diminish the utility of voice data. My work centers on evaluating the trad-off between maintaining effective anonymization and preserving the utility of speech data for its intended applications.
Previously, I worked as an AI Engineer at the French Presidential Office, Élysée Palace, where I was in charge of designing and deploying AI systems to support the work of the President of the French Republic, his Cabinet, and associated administrative teams. This work provided analytical support for decision-making and was conducted under strict constraints related to confidentiality, security, and on-premise infrastructure.
I hold a Master’s degree in Computer Science with a specialization in Artificial Intelligence and Big Data, and a Master’s degree in Industrial Engineering focused on project design and management.
Institutions: Technische Universität Berlin (Germany), INRIA (France)
Secondment: Vocapia
Supervisors: Sebastian Möller (TUB), Dorothea Kolossa (TUB), Slim Ouni (INRIA), Bernard Prouts (Vocapia), Lori Lamel (LIMSI-CNRS)
Contact: menestrel@tu-berlin.de