| Abstrakt: |
Internet usage is increasing day by day all over the world and as a result, technology is also developing to make daily life appliances as smart as possible. Millions of devices are connected using IoT technology, and the vulnerabilities of these devices are still exploitable by attackers. Having access to IoT devices through a Bot Master allows the Bot Master to attack a targeted server with these devices. To detect malicious traffic in IoT networks, there is a need for an intelligent mechanism. Although there have been many studies on the detection of botnet malware, accuracy and efficiency remain a gap. This study focuses on an automatic system that can detect botnet malware with high accuracy. A new ensemble model has been proposed in this study, known as the Extra Tree Random Voting Ensemble Classifier (ER-VEC), which is a combination of two tree-based models called Extra Tree and Random Forest. The proposed model is tested on several malicious traffic in the IoT networks datasets such as IoTID20, MedBIoT, UNSW-NB15, N-BaIoT, and ER-VEC achieving 99.99%, 99.91%, 95.64%, and 100% accuracy scores, respectively. In comparison with the proposed model, other machine learning models were also employed, and ER-VEC significantly outperformed them in terms of accuracy, precision, recall, F1-score, and error rate across all datasets. In addition, we performed K-Fold cross-validation and found that ER-VEC achieved an accuracy score of 98% and a standard deviation of 0.04±. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink