Explainable Deep Learning Models for Wireless Sensor Networks (WSNs)
Original Research | 2026 | July | Volume 1 | Issue 2 | Page 62-82
Vijay Kumar *1, Dr. Ashwani Gupta *2
Research Scholar, CSIT Department, MJP Rohilkhand University Bareilly, India. vk2007703@gmail.com
Dr. Ashwani Gupta, Assistant Professor, CSIT Department, MJP Rohilkhand University Bareilly, India. ashwani.gupta@mjpru.ac.in
Abstract
Wireless Sensor Networks (WSNs) have become an essential component of modern Internet of Things (IoT) ecosystems, enabling real-time monitoring in applications such as environmental surveillance, smart agriculture, industrial automation, healthcare, and military operations. Deep learning (DL) techniques have significantly improved the performance of WSNs by enabling accurate anomaly detection, intrusion detection, fault diagnosis, data aggregation, energy optimization, and predictive analytics. However, the "black-box" nature of conventional deep learning models limits their adoption in safety-critical and resource-constrained environments, where transparency, interpretability, and trust are essential. Explainable Artificial Intelligence (XAI) addresses this challenge by providing human-understandable explanations for model predictions without substantially compromising predictive performance. This study proposes an explainable deep learning framework for WSN applications that integrates lightweight neural network architectures with post-hoc explanation techniques, including SHAP (SHapley Additive exPlanations), LIME (Local Interpretable Model-Agnostic Explanations), and feature attribution methods. The proposed framework is evaluated using benchmark WSN datasets for intrusion detection and sensor fault classification. Model performance is assessed using accuracy, precision, recall, F1-score, energy consumption, inference latency, and explainability metrics such as feature importance consistency and explanation fidelity. Experimental results demonstrate that the explainable models achieve high classification accuracy (>97%) while providing transparent decision-making that enables network administrators to identify the factors influencing predictions. The lightweight architecture also maintains computational efficiency suitable for edge-enabled WSN deployments. The findings suggest that integrating explainability into deep learning enhances the reliability, accountability, and practical deployment of intelligent WSN systems. This approach supports informed decision-making, facilitates debugging and model validation, and promotes user trust in AI-driven wireless sensing applications. The proposed framework offers a promising direction for developing transparent, secure, and energy-efficient intelligent WSNs for next-generation IoT environments.
Keywords: Wireless Sensor Networks, Explainable Artificial Intelligence, Deep Learning, Internet of Things, Intrusion Detection, Sensor Fault Detection, SHAP, LIME, Edge Computing.
Comments