Research on Machine Learning-Based Packet-Level Device Fingerprinting for the Internet of Things

Internet of Things (IoT) devices help connect information across a variety of domains, such as smart homes, smart grids, and healthcare. However, this connectivity creates a large attack surface for cyberattacks. Fingerprinting techniques for IoT device identification can help perform asset identification during digital forensics. This research presents a solution to fingerprinting IoT devices using machine learning. The proposed solution consists of a machine-learning model based on packetlevel features that can be extracted from encrypted packets. A machine-learning model was constructed for validation, and fingerprinting was performed on a dataset of IoT devices using Zigbee. The prediction quality was found to be 93.89% accurate, with an F1 Score of 0.7538. Designed to identify not only the device name but also the device type and vendor name, the proposed framework is expected to be able to derive information about unknown devices.