Researcher in AI : Deepfake detection, Object detection and tracking

Purpose

Deepfakes pose a significant threat to the reliability of information and the security of biometric systems. The candidate will be tasked with developing automatic deepfake detection algorithms to enhance the security of our biometric solutions.

IDEMIA is recruiting researchers to enrich and strengthen its teams. The goal is to cover all aspects of scientific research from data collection to modeling, implementation, and ultimately the publication of articles or patents. The methodologies employed include Bayesian approaches, machine learning, deep learning (CNN, Transformer, GAN, Diffusion, and VLM), as well as computer vision techniques for optimized analysis of still images and automated video processing.

Key Missions

Technological Environment :

• Rapid progress is currently being made in the field of Generative AI and deepfake technology [1, 2]
• Although deepfake videos are not yet perfect, they have already caused significant harm to many individuals [3], and it is widely acknowledged that they represent a serious threat to society
• For example, deepfakes have recently been used to attempt to misinform the Ukrainian armed forces [4]
• It is therefore of utmost importance that defenses against this threatening technology remain robust and up to date
• Public information campaigns and word-of-mouth have already made the general public more skeptical about videos coming from untrusted sources
• However, more systematic and reliable defense mechanisms are required
• In the absence of strict governance regarding media provenance and distribution, deepfake detection remains the only realistic option
• Automatic deepfake detection relies on deep learning–based classification methods, whose performance largely depends on the quality and quantity of data available for neural network training

[1] Rombach et al. High-resolution image synthesis with latent diffusion models. In CVPR 2022

[2] Perov et al. DeepFaceLab: Integrated, flexible and extensible face-swapping framework. arXiv 2020

[3] https://interestingengineering.com/culture/deepfake-scam-china-concerns-ai-powered-fraud

[4] https://nypost.com/2022/03/17/deepfake-video-shows-volodymyr-zelensky-telling-ukrainians-to-surrender/

Missions : 

• Studying and developing cutting-edge algorithms, including GenAI techniques
• Optimizing algorithms
• Measuring and analyzing performance
• Conducting technology watch and reviewing the state of the art in research
• Providing expertise to analyze field-related challenges and excel in benchmarks
• Supporting operational teams in defining the best customized solutions for specific clients
• Integrating solutions across diverse environments such as cloud or on-premise servers, embedded devices, and smartphones

The scope of work will depend on the candidate’s prior experience and the potential opportunities within IDEMIA. We welcome individuals with varying levels of experience and expertise: newcomers will benefit from tailored mentorship, while seasoned professionals will find stimulating technical challenges. Our team values continuous learning and experimentation.

Profile & Other Information

Education:

• PhD or Engineering School

Technical Skills:

• Initial significant experience in image processing, computer vision, and machine learning
• Strong background in mathematics, data analysis, C, Python, and PyTorch
• Ability to handle large datasets and work in a research environment, i.e., in unfamiliar domains
• Knowledge of biometrics or document analysis would be a plus
• Fluent in spoken and written English
• An additional language would be appreciated

Soft Skills :

• Curious
• Proactive and autonomous
• Results- and solution-oriented
• Clear and persuasive communication