Edge-Based AI Framework for Anomaly Detection in IoT Networks

Abstract

The growing deployment of the Internet of Things (IoT) technology increases the complexity and vulnerability of today's networks, making effective anomaly detection a demanding necessity. Conventional security solutions in the cloud are always either too slow, too bandwidth-intensive, or inflexible for IoT. In this context, this paper proposes an edge-based artificial intelligence (AI) framework for detecting anomalies in IoT networks. The framework introduced in this work leverages edge computing to provide real-time processing and analysis of data near IoT devices, enabling fast detection of anomalous network traffic. Lightweight AI models are installed on edge nodes to monitor traffic patterns and device activities with limited computation and energy. The model is sensitive to emerging attack patterns and varying network environments to maintain stable detection performance. It is successful in distinguishing anomalies such as attacks, hijacked devices and abnormal resource behavior. The experimental results show that the edge-based architecture can achieve a lower detection latency and network bandwidth usage than centralized cloud-based solutions, whereas it maintains high detection accuracy and low false positive rates. The results show that providing an AI intelligence network in conjunction with IoT devices will be highly beneficial for securing, scaling out, and operating resiliently against emerging threats in IoT networks. This architecture offers a feasible approach to securing massive IoT deployments, such as smart cities, industrial automation, and healthcare.