Multimodal Deepfake Detection
A PhD position on multimodal deepfake detection in collaboration between ÉTS Montréal and AMIAD, co-supervised by Patrick Cardinal and Thomas Thebaud.
Start date: As soon as possible.
Topic PDF: Coming soon.
Join Us!
Research Opportunities
I am currently looking for motivated students to join a range of research projects. Explore the tabs below for PhD positions (full-time, paid positions lasting 3 to 6 years), internships (full-time, paid positions lasting 3 to 6 months), and research experiences (unpaid collaborations on projects we can develop together during your undergraduate or master’s studies). Current topics are outlined below, but I also welcome new ideas—please feel free to contact me to discuss your interests.
Six-month paid Master 2 internships at the AMIAD, starting in spring 2027, on the Polytechnique campus, Palaiseau. If you are interested in multiple topics at once, please put all the points of contact in copy so we can all meet at once.
Radio Data Simulation for Speech Processing Robustness
Description: Develop a radio data simulator and algorithms to optimize its use for training signal representation models for both audio and radio signals. The simulation must account for protocol aspects and propagation-induced perturbations.
Point of contact:
- Raphaël Duroselle: raphael.duroselle-fourcade@polytechnique.edu
Streaming Masking of Personal Information in Speech
Description: Develop a speech anonymization system capable of detecting personal information and modifying speaker words with minimal latency. The system should maintain speech intelligibility while removing personally identifiable information.
Point of contact:
- Thomas Thebaud: thomas.thebaud@polytechnique.edu
Audio Deepfake Detection Using a Database of Signatures
Description: Develop an audio deepfake detection system based on a generator signature database. The goal is to decouple the set of attacks considered by the detector from model training, allowing new generators to be added without retraining.
Points of contact:
- Thomas Thebaud: thomas.thebaud@polytechnique.edu
- Raphaël Duroselle: raphael.duroselle-fourcade@polytechnique.edu
Audio Steganography and Watermarking: Robustness, Detection, and Identification
Description: Implement and compare state-of-the-art open-source audio steganography solutions, then develop neural network-based solutions for detecting the presence of steganography/watermarking in audio recordings and identifying the method used.
Point of contact:
- Jean-François Bonastre: jean-francois.bonastre@polytechnique.edu
Agentic LLM Approaches for Scientific Paper Review
Description: Explore multi-agent approaches based on large language models for scientific paper review, with integration of text and image modalities, and verification of references using the existing "researchpool" knowledge base.
Points of contact:
- Gianni Franchi: gianni.franchi@polytechnique.edu
- Christophe Servan: christophe.servan@polytechnique.edu
Legal Document Deepfake: Document Fraud Detection in the Judicial Field
Description: Study automatically falsified documents using generative tools in the legal context. Establish a state-of-the-art of techniques for detecting fake documents, including counterfeit documents and falsified authentic documents.
Point of contact:
- Christophe Servan: christophe.servan@polytechnique.edu
Unpaid research experiences for a semester or a year, which can be rewarded with academic credits depending on your institution.
Johns Hopkins University
Research topics for Johns Hopkins University students include:
- Language identification in crowds
- Conversation analysis using large language models
- Reverberation of audio and speech signals in open and enclosed spaces
Palaiseau Engineering Schools
I welcome research project proposals from Master 1 and Master 2 students at engineering schools in Palaiseau. If you have an idea you would like to explore, please get in touch to discuss potential collaborations.