Optimized Lightweight Federated Learning for Botnet Detection in Smart
Critical Infrastructure
Abstract
In this paper, we propose an optimized lightweight Federated Deep
Learning (FDL) method for botnet attack detection in smart critical
infrastructure. First, an optimization method is developed to determine
the most appropriate combination of model hyperparameters for local Deep
Learning (DL) at the edge nodes. Then, an oversampling algorithm is
combined with the optimal DL model to improve the classification
performance when the training data is highly imbalanced, without a
significant increase in the overall computation time. Furthermore, a
feature dimensionality reduction method is used to reduce the amount of
memory space required to store the network traffic data at the edge
nodes.