The concept of RF fingerprinting is that electronic devices can be identified and authenticated through their radio frequency emissions. RF fingerprinting has been applied to enhance the security of wireless networks based on WiFi or cellular communication standards. In this paper, we apply RF fingerprinting to IoT devices on the basis of entropy based statistical features called permutation entropy and dispersion entropy. The authentication accuracy obtained using these two features is compared to other statistical features including Shannon entropy. A comparison of the obtained classification accuracy with different machine learning algorithms is performed. We demonstrate that IoT devices can be classified with great accuracy and that permutation entropy and dispersion entropy provides better accuracy than other statistical features.

BALDINI Gianmarco;  GIULIANI Raimondo;  STERI Gary;  NEISSE Ricardo; 

2017-09-05

IEEE

JRC105674

978-1-5090-5873-0,   

http://ieeexplore.ieee.org/document/8016272/,    https://publications.jrc.ec.europa.eu/repository/handle/JRC105674,   

10.1109/GIOTS.2017.8016272,