Domain Knowledge Driven Unified Architecture for IoT Device Identification using Deep Metric Representation Learning

Devices connected to the Internet of Things (IoT) are well recognized for being the root of several security issues, and as a result, they would significantly benefit from automated management. For the purpose of implementing the proper network security measures, robustly identifying devices is necessary. The performance and security of IoT devices are monitored and improved in large part thanks to device identification. Thus, methods to properly recognize IoT devices are explored. When detecting the many distinct categories of Smart home devices as indicated to differentiate devices, IoT devices confront us with both inherent difficulties and prospects. Based on important results from preceding research on comprehending the patterns of network traffic in IoT devices, an efficient machine-deep-learning IoT Network identifier approach is created in this study. In order to extract 298 distinct features of network traffic, the proposed model integrates three separate parameters such as network servers in remote, port numbers, the time between packets which is the characteristics of traffic in packet level, and the duration of the flow which is the characteristics of traffic in flowlevel. The fact that distinct devices on a network might create traffic at varying rates and this variability are considered, both during the process of training the system and while evaluating its performance. This isolates the proposed methodology from prior literature.