DDoS Attack Detection in a Real Urban IoT Environment Using Federated Deep Learning

today, alongside the opportunities provided by Internet of Things (IoT), Distributed Denial of Service (DDoS) attacks are one of the most significant attacks that target the overall availability and reliability of the network. Many researches have been devoted to propose new machine learning-based detection mechanisms. However, centralized learning models require the traffic data and learning process to be concentrated on a specific device, which leads to more computational complexity and privacy concerns. Consequently, in this paper, detection and prediction of such attacks is modeled as a distributed cooperative learning scheme, which is conducted based on federated deep learning implemented in a real smart city environment. The results compared with traditional centralized deep learning models indicate high performance and accuracy, while maintaining confidentiality of traffic data. More precisely, in terms of common learning metrics, our proposed model is capable of gaining 0.953 and 0.0369 accuracy and loss rates, respectively.