Machine Learning Based Detection of DoS Attacks in Integrated MQTT-CoAP Heterogeneous IoT Network

Internet of things is revolutionary field in number of applications which connects huge numbers of things and devices in a network. These devices are mainly heterogeneous and communicate with each other with different communication protocols like MQTT, CoAP and HTTP. It is referred as Hybrid IoT architecture. This paper explores the integration of MQTT and CoAP-based IoT network devices on a unified platform facilitated by a Software-Defined Networking (SDN) controller, leading to improved interoperability and communication efficiency within the IoT ecosystem. To enhance security, a hybrid system employing machine learning (ML) is proposed, employing Support Vector Machines (SVM), Decision Trees (DT), Naive Bayes (NB), and K-nearest Neighbor (KNN) classifiers. Trained on a comprehensive Denial of Service (DoS) dataset, these classifiers demonstrate superior performance in detecting DoS attacks, with the SVM classifier achieving an 84 % accuracy, 92 % sensitivity, and 94% specificity. The study leverages SDN controller-based OpenFlow simulation and Python-based ML classification, providing a robust architecture for IoT network security. The research suggests potential extensions using deep learning approaches to further enhance classification performance in detecting security threats.