IoT Network Device Classification Using Traffic Log by Applying Machine Learning Techniques

Network security challenges of the Internet of Things (IoT) appliances from a variety of suppliers used in wide areas, are rising quickly. Thus, the maintenance of these devices are extremely crucial to internet providers. However, it is important that devices are routinely tested for their smooth execution and for diagnostic security threats. But there are very few public datasets available with a large variety of devices for analysis of the behaviour of these devices. In this paper, we overcome these problems through the development of an effective dataset of IoT traffic aimed at enabling accurate IoT device identification and behaviour analysis. First, we built a smart environment with 24 diverse IoT devices and captured network traffic traces from this smart framework for 6 months duration. Secondly, we process traffic traces to extract packet-level features and flow-level features for the testing of our captured traffic traces. Third, we developed various frameworks for the security of smart environment automation by detection of IoT devices using ML as well as DL technologies. Lastly, we analysed the accuracy, significance, and flexibility of every machine learning and deep learning technique in offline mode as well as real-time mode. Our research opens up the opportunity for IoT accessibility, flexibility, and network security management in intelligent contexts without specialized devices or standards.