Securing Wireless Sensor Networks with an Intrusion Detection Technique Based on Grid Search Random Forest and Modified Feature Selection Technique

Sensor networks (WSNs) are fundamental components of multi-hop cyberphysical systems, incorporating stationary and mobile components. There are, however, many attacks that can disrupt the operation of WSNs and compromise security. This paper presents an Intrusion Detection System (IDS) framework that utilizes Grid Search Random Forest algorithms along with Boruta feature selection to meet these challenges. With this framework, WSN intrusions will be detected more accurately. Two benchmark datasets are used in the framework, NSL-KDD and CICIDS2017, representing different scenarios of network intrusion. Data are preprocessed through the use of Morphological Filtering and Extended Empirical Wavelet Transform, which eliminates noise and irrelevant features from the data. As a result of this preprocessing step, the model will be able to detect more accurately.