A New Approach of Optimizing Machine Learning Classification Using Intersection-weighting Feature Selection for Botnet Attack Detection

Botnets are dangerous cyberattacks and must be dealt with carefully. Previous studies have introduced Botnet detection models, but they are still not optimal and require appropriate feature selection techniques to enhance detection performance. This study proposes a feature selection technique through intersection-weighting feature analysis to optimize machine learning-based classification models. The aim is to improve the classification model's performance through feature selection analysis techniques. The novelty of this research lies in optimizing detection techniques through feature selection based on intersection-weighting feature analysis to obtain important features. Four different datasets are used in the experiment, namely NCC-2, CTU-13, NCC-1 and UNSW NB-15, and show that the Decision Tree model achieves the best average performance, with accuracy of 98.81, precision 97.23, recall 95.33, and F1-score 96.27. In contrast, the average computation time is 91.213 seconds. The proposed model helps network administrators to analyze botnet malware attacks, enabling them to identify threats earlier.