5G ABNORMAL SIGNALING DETECTION BASED ON AUTO-ENCODER

With the proliferation of 5G networks, increasing attention has been directed toward associated security risks. The N4 interface, serving as the interface between the user plane and control plane in 5G architecture, encompasses functionalities including session management and policy enforcement, and is susceptible to risks such as session hijacking. PFCP, as the application layer protocol of the N4 interface, can be effectively monitored through anomaly detection to identify abnormal behaviors within the N4 interface. Consequently, this paper proposes an autoencoder algorithm model implemented with Transformer neural networks. During the training process, the model learns the sequential characteristics of normal PFCP bidirectional flows. In the detection phase, data is processed through the encoder and decoder, and the model computes the Euclidean distance between the reconstructed data and the original data to derive an anomaly score. Additionally, this paper employs publicly available datasets to experimentally validate the efficacy of the algorithmic model in detecting PFCP traffic anomalies.