Federated deep learning for anomaly detection in the internet of things

Privacy has emerged as a top worry as a result of the development of zero-day hacks because IoT devices produce and transmit sensitive information through the regular internet. This study suggests a deep neural network (DNN) and federated learning (FL) for an IoT network as well as mutual information (MI) for an effective anomaly detection method. The suggested method is different from the conventional model by use of decentralized on-device data to spot IoT network incursions. The information is kept on localized IoT devices for model training and only modified weights are shared in the centralized FL server is an advantage of integrating FL with Deep learning (DL). It uses the IoT-Botnet 2020 dataset for evaluation. Results demonstrate the effi-ciency of the DNN-based network intrusion detection system (NIDS) in comparison to the deep learning models with improvement in the accuracy of the model and a reduction in the False Alarm rate (FAR).