Detection of IoT botnets using artificial intelligence techniques for Mirai and BashLite attacks identification
Nekunj Khanna, Usha Jain, Sushama Tanwar · 2025
The mass adoption of Internet of Things (IoT) devices has brought about substantial security concerns, with botnet attacks becoming a significant risk to the stability of networks. Traditional intrusion detection systems frequently struggle to identify advanced botnet activities, prompting the adoption of machine learning (ML) and deep learning (DL)-based techniques to achieve precise classification. This study presents an analytical comparison of two ML models (XGBoost and random forests (RF) and two DL models convolutional neural networks (CNN) and gated recurrent units (GRU) for botnet detection and classification using the N-BaIoT dataset. Our approach entails pre-processing the dataset by standardising and under sampling it and then training the models using the extracted network traffic features. The experimental findings indicate that RF achieves the highest accuracy of 99.85%, closely followed by XGBoost at 99.81%. CNN also competes in various events, while GRU exhibits the lowest performance with an accuracy of 88.29%. The results emphasise the efficiency of ensemble-based ML models in identifying botnets compared to DL architectures, underscoring their potential for practical cybersecurity applications.