Supervised Service Classification using Downlink Control Indicator in LTE Physical Downlink Control Channel
Jeong-Woo Son, Sung-Hee Lee, Minho Han · 2021 International Conference on Information and Communication Technology Convergence (ICTC) · 2021
Service classification for mobile traffic is an essential task for traffic management and service improvements. This paper proposed a supervised model based on multi-modal deep neural networks to classify mobile traffic into their services. Specifically, the proposed model is specialized to handle Downlink Control Indicator (DCI) obtained from Long Term Evolution (LTE) Physical Downlink Control CHannel (PDCCH). DCI contains control information such as Radio Network Temporary Identifier (RNTI), Resource Block (RB) assignment, and so on. Thus, it can observe which RNTI used an LTE Cell and the corresponding device used how many RBs. It is natural to regard the information in DCI as a sequential vector, the proposed model is designed with Recurrent Neural Networks (RNNs). Furthermore, dual modalities in DCI (downlink and uplink control information) are efficiently co-working by the hierarchical structure of the proposed model. With evaluations of the proposed model, we proved the efficiency of the model in the real-world data manually gathered during sixteen hours. The analysis of the experimental results suggested more problems to handle DCIs for service classification as well.