Graph-powered learning methods in the Internet of Things: A survey

Abstract

The trend of the era of the Internet of Everything has promoted the integration of various industries and the Internet of Things (IoT) technology, and the scope of influence of the IoT is developing in a wider and deeper level. With the extension of the fields involved, the in-depth progress of the IoT is facing a bottleneck. For example, the security of IoT network and software have problems that are difficult to reconcile. Graph-powered learning methods such as graph embedding and graph neural network (GNN) are expected. How to use the graph learning method in IoT is a question that has to be discussed in relation to the future of the Internet of Things. This paper comprehensively discusses related research and summarizes the progress of using graph-powered learning to promote the network anomaly detection, malware detection, IoT device management, service recommendation and other aspects of IoT. And discuss the results of using graph theory and graph-powered learning methods according to the IoT fields such as smart transportation, Industrial Internet of Things (IIoT), Social Internet of Things (SIoT), smart medical care, smart home, smart grid, and smart city. Finally, in view of the existing issues and trends, this paper proposes future research directions including city various predictions, dynamics and heterogeneity, semantic analysis, resource consumption, point cloud, digital twins, and remote sensing.

Keywords

• Internet of Things
• Graph-powered learning methods
• Graph embedding
• Graph neural network
• Graph convolution network
• IoT security