Abstract:
Botnets pose a significant security threat to IoT devices, especially with the increasing number of connected devices, leading to rapid and evolving botnet formations that present serious risks. Recent research has highlighted the use of machine learning and deep learning algorithms for detecting and categorizing botnet attacks in IoT environments, leveraging datasets like UNSW-NB15. The proposed machine learning pipeline involves exploratory data analysis, preprocessing, training, testing, and evaluation using Machine Learning Algorithms. The SVM model achieved a commendable 99.06% accuracy in classifying network traffic data as normal or malicious, despite a slightly lower F1 score of 95.52%. This model strikes a balance between accuracy and recall, effectively identifying both true positives and true negatives while maintaining a low false alarm rate of 0.93% to minimize false positives and misclassification of benign activities as harmful. This method enables the detection of botnet attacks, offering a proactive approach to prevent future assaults. By utilizing machine learning techniques, the system can detect and categorize botnet attacks efficiently. It is designed to be scalable, capable of monitoring numerous IoT devices simultaneously, and can even identify previously unknown botnet threats. The user-friendly nature of the system allows for easy integration with existing IoT devices. In essence, this proposed approach presents a robust solution for identifying and mitigating botnet attacks within the IoT landscape. Leveraging machine learning, the system offers scalable and effective detection capabilities, aiming to protect IoT devices from potential botnet threats and enhance overall security measures.
