Deep Feature Extraction Framework Based on DNN for Enhancing Mirai Attachment Classification in Machine Learning

Nowadays, the rapid growth of Internet of Things (IoT) devices has changed many parts of our lives by providing seamless connectivity and automation. However, this expansion has created new challenges and vulnerabilities, especially with regard to security. Mirai botnet, a strain of malware that has severely damaged the IoT ecosystem, is one of the prominent risks to IoT security. This research examines the impact of the Mirai botnet on IoT devices. In this study, we present a hybrid model to detect Mirai attacks. Deep neural networks (DNNs) are used to extract significant deep features, which are subsequently passed to the (Light Gradient Boosting Machine) LGBM algorithm for classification. A dataset from the Canadian Institute 2023 that included several kinds of Internet of Things assaults was used to assess the model. The model achieved excellent feature extraction and promising results with accuracy, recall, precision, and F1-score scoring up to 95 % for all. These accuracy results demonstrated the superior performance of the suggested model over competing techniques, which had scores of 71 %, 85%, and 52%, respectively, for Support Vector Machine (SVM), Light Gradient Boosting Machine (LGBM), and Stochastic Gradient Descent (SGD). Furthermore, the model outperforms the DNN model, which received a score of 65%.