Predictive Models for Smart Systems Against IoT-Based Mirai Botnet Attacks

One of the most concerning botnets nowadays is Mirai, a malware that targets distributed networks of the Internet of Things devices. To achieve the attack, specialized malware is engineered to infect and take command of these devices to carry out a Distributed Denial of Service attack on victims. Recent works have been devised to detect the Mirai botnet. The most similar works differ in approaching models or datasets from ours, leading to different levels of efficiency and effectiveness. In this paper, we attempt to leverage the power and adaptability of deep learning models, including the Convolutional Neural Network (CNN), Long-short Term Memory (LSTM), CNN-LSTM Ensemble, and Transformer. We exploit a portion of the CIC-IoT 2023 dataset which contains over 3 million samples with 46 features recording complex Botnet traces. This partial dataset is statically cleansed and then sent through the training phase. Our experimental results show promising outcomes compared to the most relevant works.