Enhanced Detection of Cyberattacks in Wireless Sensor Networks and IoT-Networks Using Powerful Stacking Ensemble

Cyberattacks are malicious activities that target computer systems, networks, or devices, aiming to disrupt, damage, or gain unauthorized access. Detecting these attacks is crucial to safeguarding critical infrastructure and sensitive data. Our proposed method utilizes a stacking ensemble technique integrating Gradient Boosting (optimized using Optuna), AdaBoost (tuned with Random Forest), and XGBoost to enhance the precision and resilience of cyberattack detection. We performed a comparative analysis of various machine learning models, such as Logistic Regression, SVM, KNN, Naïve Bayes, Decision Trees, and Random Forest, assessing their performance using accuracy, recall, precision, F1-score, and ROC AUC score. Our solution effectively classifies cyberattacks in IoT and WSN networks, delivering superior performance. We utilized the NSL-KDD, WSN-DS, and NF-ToN-IoT datasets. In binary classification, our model achieved accuracies of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9. 8 7 \%}$</tex> for NSL-KDD, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9. 5 2 \%}$</tex> for WSN-DS, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9. 9 2 \%}$</tex> for NF-ToN-IoT. In multi-class classification, the model also excelled, with 99.88 % accuracy for NSL-KDD and 99.45 % for WSN-DS, though the NF-ToN-IoT dataset's numerous classes made multi-class classification less feasible. Naive Bayes consistently underperformed across all scenarios.