Fog Node Selection for Low Latency Communication and Anomaly Detection in Fog Networks

The application of 5G in IOT (Internet of things) puts forward strict requirements for network latency. In view of the fog network scenario with large-scale IOT devices, this paper proposes a procedure of unsupervised learning to efficiently realize the requirement of low-latency communication. We propose an integrated K-means clustering and PCA fog computing design, which facilitates a new service of fog, anomaly detection. Computer simulation shows that, in the presence of large-scale path loss, shadow and small-scale fading channel, the system design with low latency needs to consider the deployment of dense fog nodes and adopt more frequency or power resources at the same time. The proposed procedure and machine learning design not only facilitates the selection of fog nodes, but also presents a new example of anomaly detection using fog nodes.