Deep Learning Based IoT Device Categorization Using Network Traffic Analysis

Internet of Things (IoT) is the subsequent stage of technological progress following the revolution of the Internet, which enables integration between the digital and the physical realm.Diverse IoT devices not only make ease of life but also introduce new types of network and data security threats, including data privacy, spoofing, and denial-of-service attacks.Accurate device identification may assist in mitigating these issues.Traditional device identifiers like media access control and internet protocol addresses are susceptible toward various types of security threats.Therefore, a new approach is required without using explicit identifiers.In this study, deep learning based a device fingerprinting approach is proposed, whilst only network traffic characteristics are analyzed to identify network-connected IoT devices.To create unique fingerprints a significant amount of network traffic (150 packets) header information is utilized, whilst fingerprints are represented as graphs.The proposed device fingerprinting model demonstrated high performance, achieving over 99% precision, recall, f1-score, and accuracy for individual IoT device identification.Achieved high performance (over