Edge-assisted federated learning for anomaly detection in diverse IoT network

The rapid expansion and increasing complexity of Internet of Things (IoT) networks have led to a heightened need for effective and adaptable anomaly detection techniques. The vast variety of devices, communication protocols over 5G and other networks, and data types present in diverse IoT environments poses significant challenges for traditional centralized methods. With the rapid increase of users in 5G networks will require drastic security measures in IoT. This paper proposes an edge-assisted federated learning approach for detecting anomalies in heterogeneous IoT networks, enabling robust and efficient performance across a wide range of devices and scenarios. Our proposed method combines the advantages of federated learning and edge computing, allowing IoT devices to collaboratively train a shared machine learning model while keeping their data local. This approach not only preserves privacy but also reduces communication overhead and latency, providing a scalable solution for large-scale IoT deployments. By incorporating edge computing, our method ensures that data processing occurs closer to the source, further improving efficiency and reducing the reliance on centralized cloud resources. We present a thorough evaluation of our edge-assisted federated learning approach, comparing it to traditional centralized techniques as well as other distributed learning methods. The results demonstrate that our approach achieves superior performance in detecting anomalies in diverse IoT environments while maintaining low latency, communication overhead over network, and energy consumption. Additionally, we showcase the adaptability of our method to various IoT network configurations and device capabilities, highlighting its potential as a versatile solution for real-world IoT anomaly detection challenges.