Day: augustus 7, 2024

By Armin Shokri Kalisa and Robbert Schravendijk - Artificial intelligence is infiltrating more and more applications. This raises privacy concerns regarding the vast amounts of data required to train these AI models. One of the proposed solutions is a framework called Federated Learning. This framework does, however, not guarantee security from attacks. This article covers how attackers can use backdoor attacks to poison the model resulting from Federated learning and what steps can be taken to make it more robust against these attacks.