Adaptive and Explainable AE-LSTM-IF Framework for Real-Time Anomaly Detection in Resource-Constrained Wireless Sensor Networks
Contributors
Prof Akhilesh A. Waoo
Keywords
Proceeding
Track
General Track
License
Copyright (c) 2026 Sustainable Global Societies Initiative

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Abstract
Wireless Sensor Networks (WSNs) are widely used for environmental monitoring, industrial automation, healthcare, smart cities, precision agriculture, and Internet of Things applications. The large number of resource-constrained sensor nodes and their continuous communication make WSNs vulnerable to abnormal behavior, network attacks, faulty measurements, and changing environmental conditions. Conventional machine-learning and deep-learning approaches can detect anomalies, but many existing methods suffer from limited adaptability, high false-positive rates, computational overhead, and poor interpretability. This paper proposes a conceptual hybrid AE-LSTM-IF framework with SHAP-based Explainable Artificial Intelligence (XAI) for adaptive and transparent anomaly detection in WSN environments. The proposed framework combines an LSTM Autoencoder for learning temporal patterns, Isolation Forest for anomaly scoring, an adaptive threshold mechanism for handling changing sensor behavior, and SHAP for explaining individual detection decisions. The architecture is designed to provide a balance among detection capability, adaptability, interpretability, computational efficiency, and real-time response. A conceptual comparison with SVM, CNN, LSTM, Autoencoder, and other approaches indicate that the proposed architecture can address several limitations of existing WSN anomaly detection systems. The paper also discusses research challenges and future directions involving TinyML, Federated Learning, Edge Intelligence, Transformer models, Graph Neural Networks, Green AI, and Digital Twin-based cybersecurity.