IoT Botnet Attack Detection and Comparison of Various Ensembler Learning Techniques/Models using Machine Learning

Abstract

The present study focuses on the IoT botnet attack detection by building an IoT network dataset and Comparison of Various Ensembler Learning Techniques/Models Using Machine Learning. The analysis involves extraction of all possible features that are directly available and the ones that can be estimated from network traffic. The resulting dataset can be applied to ensemble machine learning algorithm to detect botnet. The steps followed for this purpose include preparing a dataset, preprocessing the data, designing an ensemble machine learning model and evaluating the outcomes. In the dataset three learning models of K-nearest were tested: light gradient boost and extreme gradient boost. Comparing the proposed model with an existing IoT botnet attack detection model in terms of the number of functions used, the machine learning models used, and the accuracy obtained from the trained model, the results showed that an average accuracy of 97 per cent in the three-property model to train normal and attack to capture packets of IoT traffic. It was also found that Light Gradient Boost and Extreme Gradient Boost have an average precession of 96%, but Bagging K's closest neighbor has 95%. It was also seen that light Gradient Boost and Extreme Gradient Boost are better than Bagging K-Nearest Neighbor and that the Extreme Gradient Boost has the better recovery value than Light Gradient Boost.