[LUM#22] Towards better data protection
While data is the raw material of artificial intelligence, it can sometimes be particularly sensitive, especially in the healthcare sector. How can AI, confidentiality, and sovereignty be reconciled? For Aurélien Bellet, a researcher at the Desbrest Institute of Epidemiology and Public Health1, the solution lies in federated learning.
In order to include a large number of patients in clinical studies and thus make them more meaningful, research uses multicenter studies. This means that several hospitals or clinics are involved at the same time, sometimes even in several different countries. Advantage: this method allows for large-scale studies to be conducted on patients from a variety of social and geographical backgrounds.
Share without sharing
This type of organization also has a drawback: multicenter studies require health data from several institutions to be pooled on a single server, "which makes it impossible to maintain control over the data and could also jeopardize its confidentiality," explains Aurélien Bellet, a researcher at IDESP. How can medical research implement these collaborations while reducing the risk of sensitive information being disclosed? One solution is to share... without sharing. This is known as federated learning. "It allows data from each institution to be processed on site, without having to exchange, transfer, or transmit it, " explains the federated learning specialist.
To meet this challenge, researchers are creating learning algorithms capable of operating using locally stored data rather than data centralized on a server, as is the case with conventional machine learning methods. "It is then the intermediate results of this learning that are exchanged as and when required, rather than the data itself. This allows us to alternate between local learning and aggregation of results," explains Aurélien Bellet, who collaborates with university hospitals in Lille, Caen, Amiens, and Rouen, among others.
Democratization
To promote confidentiality and respect for medical ethics, federated learning "is part of the solution, even if it is often not enough to guarantee data confidentiality, " explains the researcher, whose team is also working with the International Commission on Information Technology and Freedom (CNIL) on the complex issue of data protection.
The advantages of federated learning extend far beyond the medical field. "It may also be of interest to companies that wish to collaborate with competitors without giving them access to certain sensitive information, " adds Aurélien Bellet, who also sees federated learning as an opportunity to make AI more accessible. "It's a form of democratization of artificial intelligence and machine learning, because it doesn't require investment in large infrastructure, paving the way for collaborative uses, for example by citizen collectives."
See also:
Aurélien Bellet's lecture on federated learning
Find UM podcasts now available on your favorite platform (Spotify, Deezer, Apple Podcasts, Amazon Music, etc.).
- Idesp (Inserm, UM)
↩︎