IoT Network Intrusion Detection Using Federated Learning

With the enhanced facilities the Internet of Things (IoT) provides, everyday life has become more convenient, enabling us to control home security, monitor patients, and even self-driving cars. While regular users find this ubiquitous nature of the IoT network convenient, malicious users or attackers view it as a vulnerable target for launching attacks. The IoT architecture’s inherent vulnerabilities in each layer create opportunities for potential security breaches. One of the immediate concerns in the various vulnerabilities of IoT networks is data privacy and security. Therefore, to tackle the issue of data privacy, we have experimented on a federated learning (FL) technique to identify IoT network intrusion using deep neural network (DNN). To implement the FL scenario, we leveraged the client–server (C-S) socket communication technique. The C-S-based FL models are validated on three IoT intrusion datasets, viz., MQTTset, N-BaIoT, and the ToN_IoT dataset. The experimental analysis demonstrates that the FL method excel the centralized learning (CL) approach, ensuring data privacy for multiple clients (Cs).