An Optimal Machine Learning Framework for the Enhancement of Intrusion Detection in Wireless Sensor Network Using Metaheuristic and RNS Technique
Ifedotun Roseline Idowu, Ayisat Wuraola Asaju-Gbolagade, Kazeem Alagbe Gbolagade · 2023
The ability of stacking to combine models was used to detect intrusions in wireless sensor networks (WSNs) by producing predictions that outperformed those of any individual model in the ensemble in a classification scenario. The implementation adopted the use of the Z-score for data standardization, the Particle Swarm optimization (PSO) algorithm for solving high dimensionality problem in intrusion datasets and the residual number system was employed for feature extraction from the reduced datasets in order to further optimize the datasets in low power domain which also helps to improve the time complexity in stack ensemble model, it was observed that Case D (Naive Bayes + Logistic Regression +KNN (Base Classifier) with Random Forest (Meta Classifier)) outperformed other models with highest classification accuracy of 97.4736% with the instances of RNS and 95.3602% without the inclusion of RNS respectively.