Attacks and classification of attacks in internet of things messaging protocols

The Internet of Things (IoT) encompasses a technological ecosystem that improves individuals daily lives by increasing productivity, comfort, health and sustainability. In addition, the Internet of Things provides a variety of benefits to many industries, including increased efficiency, productivity and cost savings. However, the spread of Internet of Things technologies has revealed many security vulnerabilities. Emerging security vulnerabilities may negatively affect users. Preventing these vulnerabilities will make the use of IoT technology more reliable. Two main contributions are made in this thesis study. First, attacks on the MQTT protocol, one of the most used protocols in IoT technology, were focused on. The negative effects of attacks on the MQTT protocol have been revealed. Secondly, machine learning techniques were studied to detect IoT attacks. An experimental environment using the Thingspeak platform, one of the IoT online platforms, was established. In this experimental environment, DDoS, Http Flood, Syn Flood, Port Scan and Udp Flood attacks were made on the IoT device. A data set was created with the data collected as a result of these attacks. Seven machine learning classification algorithms, including AdaBoost, kNN, Logistic Regression, Naive Bayes, Random Forest, SVM and Decision Tree, were applied on this created data set. As a result of these classifications, 99% accuracy was obtained with the Random Forest classification algorithm. Keywords: Internet of Things, MQTT, IoT Attacks, Machine Learning, Thingspeak, IoT Security