An Effective Detection and Classification Approach for DoS Attacks in Wireless Sensor Networks Using Deep Transfer Learning Models and Majority Voting

The Internet of Things (IoT) has been used in various critical fields, including healthcare, elderly surveillance, autonomous transportation, and energy management. As a result of its emergence, several IoT-based smart applications have been established in various domains. Wireless Sensor Networks (WSNs) are the most common infrastructure for these applications. A WSN is a network that includes many diverse sensor nodes. These nodes are scattered across large areas, collecting data and transmitting it wirelessly. These networks are subjected to a variety of security threats. Protecting WSNs against incoming threats is both necessary and challenging. This paper presents a deep transfer learning based approach for intrusion detection and classification in WSNs. To identify and categorize Denial-of-Service (DoS) attacks, we deployed several pre-trained Convolutional Neural Networks (CNNs). To improve the classification performance, the final outputs of the CNN models are combined using ensemble learning, precisely the majority voting method. We used the recent and rich WSN-DS dataset for the experiments, which includes four types of DoS attacks as well as benign samples. The experimental findings confirm the effectiveness of the suggested method, which provides an accuracy of 100%.