[visionlist] M2 internship position in Security of Deepfake Detectors Against Universal Attacks @GREYC

[Christophe Charrier]

Hi,

We are looking for a highy motivated Master’s level (M2) student or 
final-year engineering student for an internship focused on deepfakes. 
This internship will take place within the SAFE team of the GREYC 
laboratory (UMR 6072), a digital sciences research laboratory based in 
Caen, Normandy, France.

*CONTEXT*
The emergence of deepfakes and face-swapping techniques has profoundly 
transformed the landscape of visual cybersecurity. Thanks to generative 
models such as GANs and diffusion models, it is now possible to produce 
hyper-realistic videos in which one person’s face is replaced by 
another’s, in a way that is almost imperceptible to the human eye.

To counter this threat, many automatic deepfake detection models have 
been developed, leveraging deep vision architectures such as 
Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and 
spatio-temporal models. These systems aim to identify subtle artifacts 
introduced by generation algorithms (lighting inconsistencies, 
compression artifacts, local facial deformations, etc.).

However, these detectors remain vulnerable to a new class of threats: 
adversarial attacks. Such attacks consist of adding a 
quasi-imperceptible perturbation to an image in order to fool an 
artificial intelligence model. Among them, Universal Adversarial 
Perturbations (UAPs) represent an advanced form of attack, as a single 
perturbation can mislead a model across a large number of images, or 
even an entire video stream.

*OBJECTIVES*
The main objective of this internship is to evaluate the robustness of 
deepfake detection systems. To this end, the intern will design and 
assess universal adversarial perturbations (UAPs) capable of deceiving 
deepfake detectors while remaining visually imperceptible.

This 4- to 6-month internship, starting on April 1st, is intended for 
Master’s students or final-year engineering students with strong 
knowledge in deep learning and neural networks, particularly GANs, 
experience in computer vision and image processing, and solid 
programming skills (Python, TensorFlow, or PyTorch).

*SUPERVISORS*
   - Christophe Charrier, Full Professor, University of Caen Normandy – 
christophe.charrier at unicaen.fr
   - Emmanuel Giguet, CNRS Research Scientist – emmanuel.giguet at unicaen.fr

*APPLICATION*
To apply, please send an application package by email to the 
supervisors, including a CV, a cover letter, academic transcripts from 
the last two years, and any additional material that may strengthen the 
application (e.g., letters of recommendation).

*APPLICATION DEADLINE*: February 28, 2026

Best regards,

Christophe Charrier

-- 
Full Professor in Computer Science, Université de Caen Normandie

Research                    | Teaching
GREYC Lab., UMR CNRS 6072.  | IUT Grand Ouest Normandie
Campus 2, 6 bd. Mal Juin.   | Multimedia and Internet Dept.
F-14032, Caen               | 120, rue de l'exode
                             | F-50000, Saint-Lô
Tel: +33 2 31 53 81 85      | +33 2 33 77 55 10

URL :https://charrierc.users.greyc.fr/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://visionscience.com/pipermail/visionlist_visionscience.com/attachments/20260205/7a30dd51/attachment-0001.html>