Machine Learning Model for Efficient Botnet Attack Detection and Classification

Abstract

Botnets constitute a significant threat to cybersecu- rity, enabling large-scale malicious operations such as distributed denial-of-service attacks, data exfiltration, and unauthorized system access. This research presents a machine learning-based framework for the detection and classification of botnet attacks, utilizing Decision Tree, XGBoost, and Logistic Regression al- gorithms. The UNSW-NB15 dataset is employed, with distinct training and testing splits to ensure model generalization and to prevent overfitting. Feature selection techniques are applied to en- hance model performance and reduce computational complexity. Model evaluation is conducted using confusion matrices and Re- ceiver Operating Characteristic–Area Under Curve (ROC-AUC) metrics to provide a comprehensive assessment. Experimental results indicate that ensemble methods, particularly XGBoost, deliver superior performance in accurately detecting and cate- gorizing botnet traffic across various attack types. The findings highlight the effectiveness of machine learning approaches in improving the robustness and scalability of network intrusion detection systems.