Automated IoT Device Identification Based on Full Packet Information Using Real-Time Network Traffic

Swaroop Dora, Brijesh Peshvani, Ashutosh Kumar, Piyush Tiwari, S. Venkatesan, Sandeep Kumar Shukla, Rahamatullah Khondoker

Abstract: The Internet of Things (IoT) is growing and gaining popularity at a very fast rate, which is considered to be the next revolution. As its popularity increases, several challenges include unauthorized access, which may result in losing sensitive data, Distributed Denial-of-Service (DDoS) attacks aimed to make devices unavailable, and other attacks that may compromise the device. This paper focuses on device identification, which recognizes device information within the network. Device identification is achieved without having physical access to the device under consideration. Several studies have explored machine learning as an approach to passive device identification, yielding promising results. However, a standardized, automated approach to device identification without requiring the setup of external hardware devices is yet to be established. This paper uses supervised machine learning algorithms for identifying IoT device types and relies exclusively on IP addresses and port scanning techniques. The proposed model does not require any external hardware device apart from the machine where the algorithm executes or/a setup of complex network infrastructures for port mirroring. Hence, the model is readily available for use within any network. Our model achieves a 96% success rate in accurately detecting IP-based IoT devices in the network, as demonstrated by the evaluation.