Seminar (KISON): "Federated learning for data-privacy preservation"

Program

Abstract

In the era of data-driven decision-making, there is a growing need for scalable, customizable, and privacy-preserving machine learning (ML) solutions to prove the full potential of ML in real-world scenarios. Federated learning (FL) is an emerging paradigm in ML that enables collaborative model training among decentralized entities without the need for centralized data aggregation. This technique allows multiple edge devices or organizations to collectively train a shared machine learning model while keeping sensitive data localized, thereby preserving privacy. The significance of FL lies in its ability to facilitate collaborative learning among institutions without compromising data confidentiality.

Two different approaches are going to be discussed in this brief presentation. First, an introduction to evaluate Federated Learning as a solution to leverage medical data and computational resources from multiple decentralized healthcare institutions, without the need to directly share their sensitive data with a centralized server. Second, the insights of collaborative ML capabilities to ensure data-privacy preservation in an Intrusion Detection System for 5G Internet of Things (IoT) Networks.

Albert Garcia i Llagostera

INVESTIGO research assistant of the KISON research group.

*Open seminar. No registration is required.