Live traffic analysis on S1-MME interface using LSTM autoencoder

The spreading of mobile communication networks resulted in a vast amount of signaling traffic on the control plane. It is quite challenging to analyze the extensive number of message sequences, recognize patterns and identify anomalies manually. However, cutting-edge machine learning methods offer excellent tools to explore unusual events in signaling traffic to provide countermeasures in the mobile network. The paper proposes a data exploration method with an LSTM-Autoencoder based latent space model to visualize manifolds controlling the generation of control plane traffic. The model has been validated on a live LTE signaling traffic captured on the S1-MME interface.