A Combined Solution of Machine LearningMethods to Improve Botnet Detection

Nowadays, hackers use a set of infected computers to advance their destructive goals.Therefore, researchers have devoted part of their research in the field of intrusion detectionto the topic of detecting botnets. Despite multiple research efforts in this area, the accuracyof existing methods is not sufficient to provide computer user trust in networks, especiallyin the Internet of Things. In this article, a combined solution of Support Vector Machine,Random Forest, and Adaptive Boosting, all of which are machine learning methods, isproposed to improve the accuracy of bot detection systems. In this solution, after applyingmachine learning methods to the dataset, the best parameters for detection are selected.The results of the conducted experiments show an accuracy of ۹۹.۹۹% for the Bot-IoTdataset and ۹۹.۹۹% for the IoTID۲۰ dataset. Additionally, for comparison with previousworks, an accuracy of ۹۹.۹۲% has been achieved for the CTU-۱۳ dataset. In conclusion, itis evident that by utilizing a combined machine learning-based approach, very good resultscan be obtained in botnet detection.

کلیدواژه ها:

Botnet Detection ، SupportVector Machine ، AdaptiveBoosting ، Random Forest ، Bot-IoT dataset