Research On The Identification Technology Based On CSI Wireless Sensor

Recent research advancement of wireless sensing technology has made device-free interaction with WiFi-enabled the IoT environment possible.Although such smart environment greatly improves the usability by allowing the users to trigger the personalized services with certain predefined gestures,it also introduces various security problems,e.g.shoulder surfing attacks.By spoofing the gesture orders of legitimate users,the attacker could easily access the private information or services,and cause even worse consequences.Designing a secure interaction mechanism for this environment is a very challenging task.To solve these problems,this paper designs a secure interaction mechanism called SiWi by using channel state information.It gets available CSI data by commercial WiFi devices.It implements the identification by using the fine-grained multipath propagation characteristics of CSI.SiWi system consists of two parts:1.CSI-based behavior recognition.In this stage,we filter the raw CSI data by using butterworth,principal component analysis and discrete wavelet transform.Then,the denoised data is segmented according to whether the action information is included or not.Finally,we use the hidden Markov Model for the segmented data segment to identify the type of activity.2.Identification based CSI behavior recognition.This stage analyzes the three basic actions(Push,Swing,Wave)identified in the previous stage and use the Fresnel model to obtain the direction and distance characteristics from each activity.We build the activity-based identification model to get the relationship between activity and identity.Finally,a support vector machine is used to obtain the final recognition result.The results show that our system can achieve average accuracy of 93% to identify the legitimate users and 97% to resist the spoofer.