LLM-Empowered Autonomous Edge AI for Anomaly Detection and Resolution in Distributed Systems

There is a growing need for strong security measures to combat cyberattacks with increasing data consumption and reliance on interconnected networks. This is especially required at the network edge, where the data are present and first acquired. Cyber attacks such as distributed denial-of-service (DDoS) attacks and others pose risk and various other inconveniences to all stakeholder involved. The dynamic nature of these attacks and the lack of processing power at the edge require new methods to tackle cyber-attacks. This paper proposes a novel approach for network anomaly detection and mitigation that uses edge AI and the contextual understanding power of Large Language Models (LLMs). The system utilizes deep learning methods to identify cyberattacks where the models are deployed on edge devices and servers enabling quick identification of anomalies. The paper then explores the use of LLMs to automate the resolution and mitigation of these attacks in real-time, leading to lower impact by reducing downtimes and increasing availability. This research aims to contribute to the field of network security by proposing an edge AI driven solution for real-time local DDoS attack detection and the utilization of LLMs for network anomaly resolution and system adaptation.