Authentication Method for WiFi Connection of Devices Based on Channel State Information
Huan Dai, Wenhua Shi, Zelun Zhou, Jingjing Jiang · 2020
The popularity of intelligent mobile devices enables users to access various information services anytime and anywhere, which increases people's demand for communication networks. WiFi network is one of the main ways for smart devices to access the network. Compared with using cellular network, users are more willing to connect to WiFi when there is WiFi coverage, and even automatically crack the WiFi password to connect. In order to improve the access security of WiFi devices, this paper proposes an Authentication method for WiFi connection of device based on Channel State Information (CSI). The method can effectively prevent illegal users from connecting WiFi outside the legal area. When devices are connected to WiFi in different locations, it will cause different changes in CSI. According to this feature, a fingerprint database is established in each area within the WiFi signal range. Hampel filter is used to remove the noise caused by environmental factors in CSI data stream. Support Vector Machine (SVM) is used to filter the non-limited locations. However, WiFi authentication has a strong real-time requirement. SVM algorithm which only uses cross-validation to select parameters can not quickly find the appropriate parameters to adapt to the current environment. To solve the above problems, Percentile algorithm is used to improve the efficiency of RBF kernel function parameter selection, which significantly improves the real-time performance of authentication methods. The experimental results show that the accuracy of correct authentication can reach about 95%.