Iot botnet intrusion detection system using federated learning techniques

As IoT devices become more common, they’re also becoming a bigger target for security threats, especially botnet attacks. This research presents a new approach to address these risks by introducing an Intrusion Detection System (IDS) using Federated Learning (FL). Unlike traditional methods that collect data centrally, this FL-based IDS identifies botnet threats without needing a single point for data storage, which helps to keep user data private. The model is built on the N-BaIoT dataset and uses the FedAvg aggregation technique, achieving a precision of 92% and an F1 score of 0.89, a performance that is on par with centralized models. This approach not only enhances privacy but also scales better across diverse IoT networks. However, there are some challenges. High communication demands, varied data quality from different devices, and stability issues in the model suggest there’s room for improvement. Future work will focus on making communication more efficient and reinforcing the model to resist adversarial threats. This research marks a step forward in secure, collaborative IDS frameworks for IoT networks.