Research On Device Detection And Localization By Using RF-based Technologies

With the development of wireless communication technologies and the Internet of Things,various wireless devices are widely deployed resulting in a more crowded spectrum.In addition,the access security of different wireless devices.Once the wireless devices were used by criminals,the threats to personal privacy and public security can’t be underestimated.Therefore,it is desirable to effectively recognize and localize the wireless devices.Most existing work on wireless device recognition are based on Wavelet Transform,Fourier Transform or Machine Learning to identify a single device with unknown protocol.However,the diversity of the different types of devices is insufficient in current research.And there is no method to recognize and locate multiple devices when they exist simultaneously.Moving along this direction,in this paper we provide the solution to recognize and locate multiple devices with different protocols when they exist simultaneously.Specifically,the main methods are proposed: 1.For a single wireless device,the features of wireless device signals are extracted by the method of Autocorrelation and Short-Time Fourier Transform(STFT)to recognize the device;2.For multiple wireless devices,we also can extract the periodic features of signals from multiple devices by the method of Autocorrelation to recognize them.3.Our localization method is to utilize the interference of the signals to locate the wireless device through a fingerprint-based localization algorithm.In the indoor environments,we tested three different types of devices including UAV、WiFi router and ZigBee sensor.When we recognize a single wireless device,we can find that the successful recognition rate is 100% when we utilize Autocorrelation,the successful recognition rate is more than 85% when we utilize STFT to recognize the periodic signal and the successful recognition rate can reach 96.7% when we utilize STFT to recognize the random signal;When we recognize multiple devices,we can find that the successful recognition rate is 100% when we utilize Autocorrelation.When the interference between different protocol devices is used for localization,the average error of localization can be narrowed down to 1.43 m through the Gradient Boosting Regression Tree algorithm.