Privacy-Preserving Federated Transformer for IoT Anomaly Detection

Abstract ―​

The rapid expansion of the Internet of Things (IoT) has significantly transformed modern digital infrastructures by enabling seamless communication among interconnected devices. However, this growth has also introduced serious security challenges, particularly in the form of cyberattacks, data breaches, and anomalous activities within IoT networks. Anomaly detection has emerged as a critical approach for identifying such threats by analyzing patterns and deviations in network behavior. This review paper provides a comprehensive analysis of existing anomaly detection techniques in IoT systems, focusing on machine learning, deep learning, and hybrid approaches. It examines the evolution of traditional methods toward advanced intelligent models capable of handling large-scale and dynamic IoT environments. The study also explores emerging technologies such as federated learning, blockchain integration, and edge–cloud architectures, which enhance privacy, scalability, and real-time processing capabilities. Furthermore, the paper identifies key challenges, including data heterogeneity, resource constraints, lack of standardized datasets, and limited model interpretability. A critical comparison of different methodologies is presented to highlight their strengths and limitations. Based on the analysis, the review emphasizes the importance of integrated and lightweight frameworks that balance accuracy, efficiency, and security. Finally, future research directions are proposed, focusing on adaptive learning models, privacy-preserving mechanisms, and cross-domain applicability. This work aims to serve as a valuable reference for researchers and practitioners seeking to develop robust and scalable anomaly detection systems in IoT environments.