Visibility graph entropy based radiometric feature for physical layer identification

Emerging physical layer identification methods have demonstrated the capability to complement and enhance the device authentication of Internet of Things networks by exploiting the uncontrollable, unclonable, and unforgeable radiometric features resulted from randomly generated hardware imperfection in wireless devices. Multiple feature-based identification has proven an efficient and feasible approach to improving identification performance. At the same time, the lack of radiometric features effective for device identification is a major problem. Most of the existing features are derived from the view of the time, frequency, or phase domain. In this study, we explore the graph domain of wireless frame’s preambles and propose a new radiometric feature called normalized horizontal visibility graph Shannon entropy (HVGE). At first, we introduce a preprocessing consisting of sample truncation and downsampling to enable the adjustment between the computational time of visibility graph (VG) conversion and the identification performance. Secondly, we propose the calculation method of the new HVGE feature from the VG representation. Finally, an experimental study using 50 off-the-shelf wireless devices was conducted to investigate the impact of the preprocessing parameters and the effect of noise and feature combinations on the identification performance gain.