ProfilIoT: A Machine Learning Approach for IoT Device Identification Based on Network Traffic Analysis

In collaboration with SUTD + CSRC

Meidan Yair, Bohadana Michael, Shabtai Asaf, Guarnizo Juan-David, Ochoa Martin, Tippenhauer Nils-Ole, Elovici Yuval

In Proceedings of The 32nd ACM Symposium On Applied Computing (SAC’17), Marrakesh, Morocco, April 3-7 2017

Link to document

In this work we apply machine learning algorithms on networktraffic data for accurate identification of IoT devicesconnected to a network. To train and evaluate the classi-fier, we collected and labeled network traffic data from ninedistinct IoT devices, and PCs and smartphones. Using supervisedlearning, we trained a multi-stage meta classifier; inthe first stage, the classifier can distinguish between trafficgenerated by IoT and non-IoT devices. In the second stage,each IoT device is associated a specific IoT device class. Theoverall IoT classification accuracy of our model is 99.281%.

Skip to content