EDGE & FOG COMPUTING: ANOMALY DETECTION IN IOT-BASED FOG COMPUTING, DATA PROCESSING AT THE EDGE

Abstract

The proliferation of Internet of Things (IoT) devices has created unprecedented data generation at network edges, necessitating distributed computing architectures that process information closer to data sources. This research investigates anomaly detection mechanisms in fog computing environments and evaluates data processing efficiency at edge nodes. Through experimental deployment of fog computing infrastructure supporting 450 IoT sensors across smart city, industrial monitoring, and healthcare scenarios, this study examines how edge-based anomaly detection reduces latency, conserves bandwidth, and improves threat response times compared to centralized cloud processing. The findings reveal that fog-based anomaly detection achieves 87% reduction in data transmission to cloud servers while maintaining 94.3% detection accuracy for network intrusions and sensor malfunctions. Edge processing reduces average response latency from 340ms in cloud-centric architectures to 23ms in fog-enabled systems, representing a 93% improvement critical for real-time applications. However, resource-constrained edge devices face challenges implementing complex machine learning models, with detection accuracy declining to 89.7% when model complexity exceeds device computational capabilities. The research identifies optimal fog node placement strategies, lightweight anomaly detection algorithms suitable for edge deployment, and hybrid edge-fog-cloud architectures balancing processing distribution. These findings have important implications for designing scalable IoT systems requiring real-time anomaly detection while managing bandwidth constraints and ensuring data security at network edges.