Machine Learning Approaches for Anomaly Detection in IoT Networks: A Survey
Kimsreng Lim, Sophea Prum · 2025
The exponential growth of IoT devices across health-care, smart cities, and industrial automation has heightened security challenges, rendering traditional detection systems in-adequate for dynamic, heterogeneous networks. This survey provides a comprehensive review of machine learning techniques for anomaly detection in IoT networks, encompassing traditional methods, deep learning, and federated learning. We critically evaluate their effectiveness, focusing on IoT-specific challenges such as resource constraints, scalability, and concept drift. This survey also discusses recent advances such as lightweight machine learning models and privacy-preserving methods like federated learning, which show promise for improving deployment in IoT environments. The paper expands the dataset review with recent, underrepresented datasets tied to emerging IoT technologies and connects future research to trends like 5G integration and privacy preservation. By addressing practical trade-offs and outlining open research questions, this work offers actionable insights and a forward-looking roadmap for securing IoT systems.