Dimensional Feature Reduction for Detecting Botnet Activities

Rising number of devices linked to the internet has made computer network security crucial. Those devices may be compromised, forming botnets, one of the most severe threats to network security due to their unique characteristics. An in-depth analysis of various processes, including feature extraction, is required to develop a botnet detection model with reliable performance. In this system, feature extraction is one of feature engineering, which is part of the data pre-processing. To find the best approach, we analyze the impact of feature extraction using dimensional reduction with four techniques: Principal Component Analysis, Truncate Singular Value Decomposition, Factor Analysis, and Fast Independent Component Analysis. The feature extraction results are brought to the classification stage to analyze their impact using several machine learning algorithms such as k-NN, Decision Tree, Random Forest, Naive Bayes, and Logistic Regression. Using the CTU-13, NCC-1, and NCC-2 datasets, it is found that dimensional reduction is suitable with k-NN but not recommended for a tree-based machine learning algorithm.