Publication

Abstract : The combination of several Machine Learning and Deep Learning techniques has been spurred by the need to address security breaches inside an Internet of Things (IoT) focused environment. This research presents a novel way to solve the challenge of classifying normal and abnormal attacks on the Domain Name System (DNS) protocol. The proposed method involves the use of a hybrid model that combines Convolutional Neural Networks (CNN) with Principal Component Analysis (PCA). The methodology begins by transforming nominal features into numerical data as part of the preprocessing stage. The quantitative data is subsequently subjected to PCA in order to identify features, reducing the dimensions of the dataset by separating the most important properties. Following this, the data is inputted into the CNN with the objective of detecting and categorizing anomalous behaviors inside the IoT ecosystem. The effectiveness of the hybrid model was assessed by employing the IoTID20 dataset. The model exhibited exceptional performance in terms of accuracy, recall, F-Score, precision, and ROC metrics, surpassing those of existing detection methods. Significantly, the suggested framework not only improves security measures but also tackles privacy concerns and strengthens the maintainability of IoT-based systems.