A Network Intrusion Detection System Based on Deep Learning Models in IoT Systems

The Internet of Things (IoT) has a vital role in the lives of people today. However, as the use of IoT devices becomes more widespread, there is a growing concern about security threats, like botnet attacks. Therefore, the use of inclusive solutions is required.   Intrusion Detection Systems (IDS) can detect and mitigate attacks on IoT devices by analyzing network traffic and device behavior. This paper proposes an IDS that uses Deep Learning (DL) techniques. It is based on an ensemble learning model that employs diversity and F۱-score as a performance metric to select the best transfer learning models. It also proposes ۲۰ individual and hybrid DL models, including Convolution Neural Networks (CNN), Recurrent Neural Networks (RNNs), and Deep Neural Networks (DNN), to detect and classify regular and botnet attack classes. The proposed IDS engages a feature engineering method to reduce unnecessary computation. The Bot-IoT dataset used in this paper contained DDoS, DoS, Reconnaissance, and theft attack labels. The proposed IDS was compared with existing methods using the Bot-IoT dataset. Experimental results disclose a high performance of the proposed model for detecting and classifying various attack and regular labels.

کلیدواژه ها:

Internet of Things ، Intrusion Detection System ، Botnet Detection ، Deep Learning ، Feature Engineering