WiPass: PIN-free and Device-free User Authentication Leveraging Behavioral Features via WiFi Channel State Information

Abstract

As an essential way of human-computer interaction, user authentication has attracted widespread attention in recent years. Traditional user authentication methods include Personal Identification Numbers (PINs) and biometric technology. However, the PINs are easily leaked to others, and biometric technology requires specialized equipment. Different from traditional user authentication methods, in this paper, we use widely deployed WiFi infrastructure to achieve user authentication, and propose WiPass, which is a PIN-free and device-free user authentication leveraging behavioral features via WiFi Channel State Information (CSI). The key idea is to explore personalized behavioral information captured by WiFi CSI to identify different users. Concretely, we first proposed a data visualization method to visualize CSI data as a set of time-series images to preserve the behavior information of keystrokes. Secondly, all these images jointly input into a Convolutional Neural Network (CNN) for feature learning, and the obtained 256-dimensional deep behavior features are used to learn a linear support vector machine classifier. To demonstrate the effectiveness of WiPass, we built a prototype of WiPass on low-cost commodity WiFi devices and verified its performance in three different real environments. The empirical results show that WiPass achieved an average of 90.5% authentication accuracy, 7.5% false acceptance rate, and 6% false rejection rate for 11 participants in three real environments.