Federated Learning (FL) presents a promising approach to collaborative model training thatmaintains data privacy by avoiding the centralization of data. This study investigates theapplication of FL, particularly Horizontal Federated Learning (HFL) and Vertical FederatedLearning (VFL), within the context of Intrusion Detection Systems (IDS) for the Internet of Things(IoT). We developed a deep learning-based Convolutional Neural Network (CNN) model andevaluated its performance across three setups: Centralized Machine Learning, HFL, and VFLusing the FEDn library for two data sets, UNSW\_NB15 open source data set and IOT dataprovided by Uppsala University researchers. The centralized model served as a baseline toevaluate the effectiveness of federated methods. HFL demonstrated robust performance,balancing precision and recall, while VFL faced several challenges related to implementation,particularly due to the novelty of integrating VFL into the FEDn library. Despite these challenges,significant progress was made in establishing a pipeline for VFL within FEDn, providing afoundation for future research. The comparative analysis of these approaches highlights thepotential of federated learning to enhance the reliability and privacy of IDS in IoT environments.