| Abstrakt: |
The Internet of Things (IoT) has various applications in practice, such as smart homes and buildings, traffic management, industrial management, and smart farming. On the other hand, security issues are raised by the growing use of IoT applications. Researchers develop machine learning models that focus on better classification accuracy and decreasing model response time to solve this security problem. In this study, we made a comparative evaluation of machine learning algorithms for intrusion detection systems on IoT networks using the DS2oS dataset. The dataset was first processed for feature extraction using the infogain feature selection approach. The original dataset (12 attributes), the dataset (6 attributes) produced using the info gain approach, and the dataset (11 attributes) obtained by eliminating the timestamp attribute were then formed. These datasets were subjected to performance testing using several machine learning methods and test choices (10-crossfold, percentage split). The test performance results are presented, and an evaluation is performed, such as accuracy, precision, recall, and F1 score. According to the test results, it has been observed that 99.42% accuracy detection rates are achieved with Random Forest for IoT devices with limited processing power. [ABSTRACT FROM AUTHOR] |
