Informations générales
Intitulé de l'offre : Post-Doc (M/F) in Multimedia Data Security - Image Obscuration (H/F)
Référence : UMR5506-WILPUE-004
Nombre de Postes : 1
Lieu de travail : MONTPELLIER
Date de publication : mercredi 2 avril 2025
Type de contrat : Chercheur en contrat CDD
Durée du contrat : 12 mois
Date d'embauche prévue : 1 juin 2025
Quotité de travail : Complet
Rémunération : From €3,021.50 gross per month depending on experience
Niveau d'études souhaité : Doctorat
Expérience souhaitée : Indifférent
Section(s) CN : 07 - Sciences de l'information : traitements, systèmes intégrés matériel-logiciel, robots, commandes, images, contenus, interactions, signaux et langues
Missions
An increasing number of methods are being developed to generate obscure images that protect the privacy of users by guaranteeing a certain level of visual security, and in particular the confidentiality of such multimedia content [Hao et al., 2020]. An image can be rendered obscure by blurring, pixelation or selective encryption, for example. Objective: The main aim of this work will be to develop new methods for generating obscure images by selective encryption that are robust to all types of attack and that would enable the content of a protected image to be recognized [Tanaka et al., 2022]. To achieve this, different types of attack will be analyzed to understand how recognition is possible.
Activités
- First, we analyze denoising attacks. Even if classical metrics such as PSNR are not convincing after a denoising attack, we find that both the human visual system and a deep neural network classifier are able to recognize the content of denoised dark images. The parameters of encryption systems can then be strengthened to protect against this type of attack. Analyses of possible reconstruction intervals for each element (e.g. pixel) will be developed.
- We will exploit conditional generative adversarial networks to develop new methods for generating obscure images that guarantee their visual confidentiality.
Compétences
- Good command of:
1) image processing and analysis
2) statistical AI: CNN, DL, GAN, etc.
3) multimedia security
Programming in C/C++ and Python.
Expected results: The aim is to develop new image encryption and obscuration methods that guarantee confidentiality and are robust to all types of sufficient attacks aimed at recognizing content.
Contexte de travail
As part of the PEPR Cybersecurity COMPROMIS project: https://www.pepr-cybersecurite.fr/
Le poste se situe dans un secteur relevant de la protection du potentiel scientifique et technique (PPST), et nécessite donc, conformément à la réglementation, que votre arrivée soit autorisée par l'autorité compétente du MESR.
Contraintes et risques
Laboratory in ZRR, FSD authorization required.