Domain Based Network Intrusion Detection System For Iot

Ensuring security of computer networks is crucial in today’s interconnected world. By identifying and reducing security risks, intrusion detection systems (IDS) play a crucial part in protecting network infrastructures. While machine learning (ML) approaches have improved IDS accuracy, there is a lack of generalized frameworks that can be implemented across different datasets, hindering practical deployment. Additionally, the disconnect between ML models and domain knowledge has reduced the explainability of IDS systems. To address these gaps, this research work designed an automated IDS framework that enhances scalability and interpretability along with real time inference. The framework focuses on automation of preprocessing, feature selection and the domain feature extraction using different algorithms. Additionally, edge deployment in improving intrusion detection for IoT devices. The results are evaluated using five Machine Learning models and contrasted using an automated and a manual feature selection procedure for each of the 10 attacks. The system concentrates on three categories of assaults for deployment purposes: Distributed Denial of Service (DDoS), Denial of Service (DoS), and normal traffic. To classify these assaults, a logistic regression model is trained and the prediction is displayed using Streamlit UI.