A Hybrid Framework Using CNN, DNN and RF for Intelligent Intrusion Detection in Wireless Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) are essential in a variety of applications, ranging from environmental monitoring to smart and connected cities to industrial automation and control. Given their resource-constrained nature and that WSNs can be deployed in a non-physical secure manner, they are particularly vulnerable to a host of cyber-attacks. Consequently, this study proposes A Hybrid CNN+DNN+RF Framework for Intelligent Intrusion Detection in Wireless Sensor Networks (WSN-DS). Our hybrid approach utilized the feature extraction capabilities of Convolutional Neural Networks (CNNs) and Deep Neural Networks (DNNs) to automatically extract the spatial and non-linear features from the WSN traffic data and subsequently classify these features using a Random Forest (RF) ensemble model to enhance accuracy, robustness, and interpretability. The system presented in this study was trained and assessed on WSN-DS training and test dataset which is the benchmark dataset for intrusion detection in WSNs. The experimental findings of this study show that the hybrid approach surpassed CNN, DNN, and conventional machine learning classifiers with regard to detection quality, along with statistically significant improvements in accuracy, precision, recall, and F1-score.  In addition, the framework also achieves acceptable false positive rates, provides greater reliability, and is therefore well suited for implementation where resources are limited, especially for real-time intrusion detection systems in a WSN environment.  This study demonstrates the promise of hybrid deep learning–ensemble methods to improve the reliable and secure operation of intelligent WSNs.