U-Plane-based Two-level Anomaly Detection Scheme for Large Scale 5G-IIoT: An Open-source approach for Aether Onramp
Yuxuan Shi, Qianqian Pan, Akihiro Nakao · 2024
The rapid expansion of the Industrial Internet of Things (IIoT) has presented challenges in system management, especially in large-scale IIoT systems with a large swarm of devices. An anomaly detection task is particularly challenging in a Radio Access Network (RAN) that contains numerous devices. The existing RAN-based methods have the following issues: 1) Most existing methods are based on Control Plane (C-Plane) Key Performance Indicators (KPIs) and lack the capability to process User Plane (U-Plane) features. 2) Existing research is mostly device-based, while application-level identification and anomaly detection are hard to perform without U-Plane features. 3) There are currently few studies and solutions for open-source mobile system approaches for anomaly detection. This paper proposes a two-level anomaly detection scheme for large-scale 5G Industrial Internet of Things (IIoT) networks based on the open-source RAN system Aether Onramp. The proposed scheme utilizes unsupervised machine learning to detect anomalies at both device and application levels, leveraging U-Plane traffic data. The scheme achieves an F1-score of 97%, a 13% improvement over existing methods, and a true positive detection rate of 64%, indicating its effectiveness in detecting anomalies. The proposed scheme has potential applications in network management and security for IIoT systems.