The rapid development of networking technologies has chang-ed how we live and allows us to grow as people and society work together. However, as we depend increasingly on these technologies, significant security worries exist, especially about people hacking into network data and intrusion. This study introduces a Federated split Intrusion Detection System (FIDS) that utilizes federated learning techniques to effectively identify, categorize, and handle network intrusions while reducing the occurrence of false positives in non-identically and independently distributed (non-IID) data sets. The system employs a Convolutional Neural Network (CNN) combined with a Long Short-Term Memory (LSTM) model, which has been trained using the CICIDS2017 dataset to reach a high level of accuracy in detection. Federated split learning offers robust privacy preservation and computational efficiency benefits; this strategy decreases the susceptibility to reverse engineering attacks, provides protection against such assaults, and enhances privacy. The system proposed in this work utilizes federated split learning techniques to protect sensitive data and ensure privacy. It is a robust and privacy-focused solution for detecting intrusions in IoT networks.