Federated Learning in the Wild: A Comparative Study for Cybersecurity under Non-IID and Unbalanced Settings
The episode discusses a paper by Doriguzzi-Corina et al. comparing Federated Learning methods for cybersecurity under non-IID and unbalanced settings, focusing on spotting DDoS attacks without private data. The hosts discuss the experimental setup using a Kubernetes testbed and conclude that adaptive learning algorithms like FLAD perform better than standard methods.