Machine Learning based IoT-BotNet Attack Detection Using Real-time Heterogeneous Data

Security is a major concern for Internet of Things (IoT) devices. Due to their vulnerabilities, these devices are an easy target for unauthorized access. Traditional security mechanisms are not easily deployed on these tiny devices. Therefore, machine learning-based algorithms are proving to be key enablers in IoT attack detection. In this paper, we critically investigate the efficacy of classification algorithms; Logistic Regression (LR), Gaussian Naive Bayes (GNB), Decision Trees (DT), Random Forest (RF), K-Nearest Neighbors KNN), and Extreme Gradient Boosting (XGB) for malicious attack detection in an IoT network flow traffic. A real-time heterogeneous dataset is used in this study. The data is preprocessed before experimentation. In addition, a feature selection method is implemented to identify the most important features. Furthermore, we evaluate the classifiers using accuracy, precision, recall, and Fl-score. The results on binary classification are comparable with the state-of-the-art in this domain. The tree-based classifiers outperform LR and GNB with 99% accuracy and a 0.99 Fl-score. It is observed that the LR and GNB are not the best choices for multiclass classification. However, DT, KNN, RF, and XGB provide an accuracy of 9S% and an Fl-score of 0. 9S. We also analyze the impact of reducing the number of features on accuracy and training time.