Anomaly-Aware Intrusion Detection Architecture for WSNs Leveraging Hybrid ML Ensembles

Abstract

Wireless sensor networks (WSN) are essential for several monitoring applications; Yet they are prone to security concerns, including unauthorized approach, attacks and other harmful actions that could endanger their reliability. Accepting the intrusion detection systems (IDS) is necessary for timely identification and response to these risks. Multiple data sets such as KDD Cup, NSL KDD, UNSW-NB 15 and AWID data are often used to train and evaluate IDS models. The choice of features is an critical manner for optimizing the electricity of the version, with techniques consisting of selectkbest used with anova F-test that offers significant reduction of functions and improved accuracy. This article examines the implementation of the stacking strategy with both bagging with random forest and boosting the algorithms of the choice tree the usage of these facts units and feature choice techniques. This approach shows great accuracy in all evaluated records units and presents a long lasting hassle solution provided by using security threats in WSN. The finding emphasizes the performance of the file technique to improve IDS overall performance for good sized safety in WSN.