A Survey on Anomaly Detection in IoT Networks: From Classical Methods to AI-Powered Solutions

IoT networks enable seamless connectivity but face security challenges due to their scale and complexity. This survey reviews anomaly detection techniques, including machine learning, deep learning, and statistical methods, analyzing their effectiveness, scalability, and limitations. More recent studies on edge and fog computing systems aim to enable real-time detection with lower latency. Most research focuses on centralized anomaly detection, making it ideal for resource-constrained IoT scenarios. The findings also show that graph neural networks (GNNs) and federated learning are promising approaches for scalable and privacy-preserving anomaly identification in large IoT networks. Despite these advances, many unanswered questions remain in areas such as data imbalance, the interpretability of ML models, and processing constraints in IoT devices.