A Productive Feature Selection Criterion for Bot-IoT Recognition based on Random Forest Algorithm

For IoT security to function properly, it is necessary to identify anomalies and suspicious activities in the Internet of things (IoT) network in order to keep an eye on things and stop undesired traffic flows in the IoT network. A large number of machine learning (ML) approach models have been suggested by many scholars to restrict fraudulent traffic flows in the Internet of Things network in order to achieve this goal. However, as a result of insufficient feature selection, several machine learning models are vulnerable to misclassifying mainly malicious traffic flows. Nonetheless, a key topic that should addressed in greater depth, and that is how to choose useful features for reliable mischievous traffic identification in an Internet of Things network, which is now being investigated. A framework is designed in order to deal with the issue. After developing and designing a novel feature selection metric strategy that relies on wrapper method to precisely filter the features, we then utilised a random forest algorithm for harmful traffic detection in an Internet of Things network to identify malicious traffic. On the basis of the Bot-IoT database, we assess the effectiveness of our suggested strategy. The examination of exploratory findings proofed that our suggested strategy is efficient and can produce output in excess of 96 percent of cases.