Intelligent anomaly detection systems in 5G networks: synthesis of AI and edge computing

Abstract

The study examines the use of AI-based anomaly detection in 5G networks, namely, the application of edge computing to mitigate security issues arising from large device densities. With the advent of 5G and decentralized architectures, conventional cloud-based security models cannot rely on high latency and bandwidth overhead. In this research, the goal is to summarize localized AI inferences on edge nodes to detect network anomalies in real time. The proposed framework uses deep learning (DL) models, namely recurrent neural networks (RNNs) and autoencoders, which are deployed directly on the edge layer. The system processes 5G traffic metadata on edge devices and uses a distributed real-time detection architecture, so there is no continuous backhaul to the core network. Findings from the implementation indicate that performance improvement is substantial, with the average detection rate for sophisticated cyber threats exceeding 98.5%. In addition, AI and edge-computing synthesis reduced detection latency by 40-55% compared to centralized cloud models, ensuring faster response times to critical network failures.