Federated learning (FL) is an approach to machine learning where a model is trained collaboratively across multiple decentralized devices or servers, while keeping the training data localized. This decentralized method enhances privacy, as sensitive data remains on the originating devices. However, federated learning also introduces unique security challenges and vulnerabilities that must be addressed to ensure the integrity, confidentiality, and availability of the model and data.
Our research in this context focuses on model authentication and mechanisms that include blockchain as a means of preventing attacks, with a particular focus on data poisoning attacks.
You can start discovering our research by looking at the following manuscript: