PhD Position on Risk Evaluation of AI Models in Healthcare

This doctoral position focuses on assessing the privacy and security risks associated with artificial intelligence models trained on sensitive healthcare data. Hosted by Inria Grenoble Rhône-Alpes and conducted in collaboration with AP-HP's digital services, the project investigates model anonymization, end-to-end attacks, and countermeasures for text and multimodal clinical data. The successful candidate will conduct state-of-the-art research, develop robust evaluation frameworks, and help secure health data sharing across hospital networks. Applicants are expected to hold a Master's degree or equivalent and possess strong backgrounds in machine learning, natural language processing, or security.

• Conduct a comprehensive literature review on privacy risks and data protection methods for machine learning and natural language processing models.
• Design and execute end-to-end privacy attacks against AI models trained on sensitive medical records.
• Develop novel defensive countermeasures and anonymization techniques to facilitate safe model sharing between hospitals.
• Integrate and scale proposed methods within actual health data warehouse environments such as AP-HP.
• Write scientific research articles and present findings at international academic venues.