Advanced Fault Detection in WSNs: A Metaheuristic-Driven Deep Learning Approach to Enhance Quality of Service

Abstract

Wireless Sensor Networks (WSNs) face significant challenges in fault detection, which directly impacts the Quality of Service (QoS) in dynamic environments. This research presents a novel framework to address these challenges, integrating a Dynamic Noise Filtering (DNF) technique with adaptive thresholding for effective noise removal while maintaining critical data integrity. The Rank-Based Whale Optimization Algorithm (RWOA) is employed for feature selection, optimizing model performance, and minimizing computational complexity. The core of the framework, the Hierarchical Attention-Based Deep Learning (HADL) model, leverages temporal convolutional layers, LSTM units, and hierarchical attention mechanisms to capture both short-term and long-term dependencies, resulting in exceptional fault detection accuracy. The proposed method demonstrates outstanding performance on the WSN-DS dataset, achieving precision, recall, F1 scores, and an AUC of 0.99 or higher for all fault classes. Comparative analysis reveals the superior performance of the framework in terms of accuracy, sensitivity, specificity, and computational efficiency. The approach not only improves fault detection but also enhances network reliability, reduces false alarms, and extends the operational lifespan of WSNs. This research offers a scalable solution for mission-critical applications, such as healthcare, environmental monitoring, and industrial automation, with potential for further enhancement through real-time deployment, multi-modal datasets, and hybrid optimization techniques.