ScCEE-IDS: A Federated Learning-Based Intrusion Detection System for Smart Cities with Cloud-Edge-End Integration

With the development of smart cities, the number of Internet of Things (IoT) devices has surged dramatically. Due to cost and performance limitations, many IoT devices inherently lack security measures. These IoT devices are widely distributed and typically connect to cloud servers through edge nodes, forming a multilayer Cloud-Edge-End architecture. However, existing Intrusion Detection Systems (IDS) for this scenario mainly focus on distributed characteristics, ignoring the data disturbances caused by IoT devices mobility. This leads to performance degradation of IDS in dynamic network environments and with changing data distributions. To address the above issue, this paper proposes an unsupervised IDS based on Federated Learning (FL), named ScCEE-IDS. This system takes into account the data disturbances caused by IoT devices mobility and is suitable for the Cloud-Edge-End collaborative architecture in smart cities. We adopt a cross-layer distributed deployment approach, where a large number of IoT nodes are responsible for collecting and preprocessing network data, while edge nodes serve as FL clients, collaboratively training with dynamic data from different sources. This effectively reduces the volume of data transmission in the network while ensuring model generalization. Experimental evaluations demonstrate that ScCEE-IDS can maintain stable detection performance in the face of dynamic data in complex networks and outperforms similar methods.