XAI-Driven Federated Learning Framework for Energy-Efficient, Secure and Real-Time Anomaly Detection in IoT-Driven Wireless Sensor Networks

Abstract

The rapid growth of IoT-driven Wireless Sensor Networks (WSNs) has enabled extensive real-time monitoring and data collection across diverse applications, from smart cities to industrial automation. These networks face critical challenges such as energy constraints, data security, and timely detection of anomalies caused by faulty or malicious nodes. Existing centralized machine learning approaches often compromise privacy and are unsuitable for resource-constrained WSN environments. To address these issues, this research proposes an Explainable Artificial Intelligence (XAI)-enabled Federated Learning (FL) framework for energy-efficient, secure, and real-time AD in IoT-driven WSNs. The methodology integrates federated learning to collaboratively train models across distributed sensor nodes without sharing raw data, ensuring privacy preservation, while XAI techniques provide transparency and interpretability of AI-driven decisions, allowing network administrators to understand and trust AD outcomes. This proposed framework also incorporates energy-aware routing and relay node optimization strategies to reduce power consumption, prolong network lifetime, and maintain real-time responsiveness. The primary objectives include enhancing AD accuracy, minimizing energy usage, preserving data privacy, and improving network reliability. Simulation results and performance evaluations are expected to demonstrate significant improvements in energy efficiency, detection speed, and interpretability compared to existing approaches, providing a robust and intelligent solution for next-generation IoT WSN management