Abstract

The majority of IoT devices are constructed based on wireless sensor node technology; therefore, wireless sensor networks (WSNs) can be considered as the foundation of the Internet of Things (IoT). Denial of service (DoS) attack is one of the dangerous attacks that WSNs are susceptible to. These attacks have the potential to impair network performance and operation. The accuracy and effectiveness of traditional methods that are used for identifying DoS assaults in WSNs are frequently lacking. This paper aims to develop a detection model based on machine learning to detect DoS assaults in WSNs. Two layers of enhancement are employed to enhance the proposed model, developing a balanced dataset and utilizing an appropriate feature selection technique. Compared to several existing approaches, the proposed model, which is based on the decision tree (DT) classifier, has achieved a high classification accuracy rate with minimum overhead. It attained classification accuracies of 100%, 99.2%, 99%, and 99.6% to detect flooding, blackhole, grayhole, and scheduling attacks, respectively. Actually, the proposed model can meet WSN limitations and constraints because it uses a lightweight classifier and employs an adequate feature selection approach that decreases computational overhead and improves performance. In conclusion, this research article offers two main contributions: it introduces an effective model for detecting DoS assaults and develops a balanced dataset (ROS-WSN-DS) using a random oversampling technique, which significantly improves the classification accuracy and performance.