Botnet Attack Detection in IoT Using Machine Learning.

Area of Science:

• Cybersecurity
• Machine Learning
• Internet of Things (IoT)

Background:

• The proliferation of Internet of Things (IoT) devices has led to a rapid increase in sophisticated cyberattacks, particularly botnets.
• These botnets pose significant risks by disrupting IoT networks and services, necessitating advanced detection and classification methods.
• Existing research increasingly focuses on Machine Learning (ML) and Deep Learning (DL) for identifying and categorizing botnet threats in IoT environments.

Purpose of the Study:

• To propose and evaluate machine learning models for binary classification of botnet attacks within the IoT ecosystem.
• To address the challenge of class imbalance in cybersecurity datasets using data augmentation techniques.
• To establish a comprehensive machine learning pipeline for robust botnet detection.

Main Methods:

• Utilized the publicly available UNSW-NB15 dataset for training and testing machine learning models.
• Implemented the SMOTE-OverSampling technique to effectively resolve class imbalance issues within the dataset.
• Developed and evaluated a machine learning pipeline encompassing exploratory data analysis, data preprocessing, and model implementation.

Main Results:

• Compared the performance of Decision Tree, XGBoost, and Logistic Regression models for botnet classification.
• Evaluated models based on key performance metrics including accuracy, F1-score, precision, and recall.
• The Decision Tree model demonstrated superior performance, achieving 94% test accuracy.

Conclusions:

• Machine learning models, particularly the Decision Tree, show significant potential for accurately classifying botnet attacks in IoT environments.
• The proposed pipeline, including data balancing techniques, provides a solid framework for developing effective IoT security solutions.
• Further research can build upon these findings to enhance the resilience of IoT systems against evolving cyber threats.