A Precise Flow Representation for Autonomous IoT-Devices Reconnaissance
Govinda M. G. Bezerra, Tadeu Nagashima Ferreira, Diogo M. F. Mattos · 2022
Devices from the Internet of Things increasingly mediate a significant number of essential everyday activities. IoT devices empower homes, industries, and offices, monitoring, sensing, and acting ubiquitously and stealthily. However, each device produces a network fingerprint that leaks information about users' behaviors and routines. This paper proposes a flow representation method for precise recognition of different types of IoT Devices. Our proposal relies on a tensor representation of the network flows to retrieve spatial and temporal correlation of flows. We show that our proposal achieves up to 99% precision on classifying IoT network flows using machine learning algorithms, such as Convolution Neural Networks, Recurrent Neural Networks and boosted decision trees.