Characterizing 5G Core Stability under Dynamic UE Conditions via Outlier Density Analysis

The dynamic nature of User Equipment (UE) life-cycle events, such as registration and de-registration, imposes significant stress on the 5G Core Network (5G CN), potentially leading to performance degradation and failures. While prior work established methods for proactive failure prediction based on control plane metrics, the direct impact of UE churn on system stability requires a deeper investigation. This paper extends previous research by characterizing the correlation between UE registration/de-registration cycles and the density of statistical outliers in network metrics. Through a series of controlled experiments simulating UE churn, we demonstrate that outlier density serves as a sensitive proxy for systemic instability and operational stress. Our findings reveal a strong, predictable pattern of correlation between the number of active UEs and outlier density, allowing us to establish a signature for benign, high-load conditions. As UEs are registered, the outlier density increases in a characteristic way, and as they are de-registered, it returns to a stable baseline. This research validates the utility of outlier density for characterizing the system’s response to dynamic UE activity, which is a critical step in distinguishing normal operational stress from anomalous patterns that precede network failures. This provides a more nuanced understanding of 5G CN behavior and enhances proactive failure prediction mechanisms by offering a simple and direct way to detect the state of the system and adapt to more robust operational configurations without relying on complex prediction algorithms.