A wireless router fingerprint identification method based on CNN

Aiming at the problem of high technical content and complex technical means of fingerprint feature extraction of wireless devices, under the premise of constant wireless space channel state, a method based on convolution neural network (CNN) for automatic classification of wireless routers is presented to solve the difficult problem of fingerprint extraction. The main work of the paper is to design and implement a method for identifying wireless routers through receiving and processing multiple-input multiple-output (MIMO) signal amplitudes. This method collects channel state information of wireless routers, pre-smooths and denoises the amplitude data of channel state information, and then use the pre-processed amplitude data as the fingerprint feature of the device, and finally is classified and identified by the machine learning algorithm. The experiment used CNN to classify and identify 10 commercial wireless routers with an accuracy rate of over 96%, and proved that using CSI to identify wireless routers is feasible.