Continuous User Authentication using WiFi

As the need for information security protection continues to increase, contactless and continuous user authentication solutions are gaining attentions. However, most of the existing schemes require the user to perform specific actions, which makes them inconvenient for practical usage. In this paper, based on widely deployed WiFi devices, we propose a framework to capture the uniqueness of human breath to achieve continuous user authentication. To achieve this, we first extract breath pattern from WiFi signal through a weighted combination of signals on different antennas and subcarriers. Then, we obtain time-domain morphological features, frequency-domain wavelet features and dynamic differential features to capture the uniqueness of the breath pattern. Finally, we utilize Gaussian Mixtures Models to achieve continuous user authentication. Extensive experiments on 24 users in 4 environments demonstrate that the proposed framework can robustly authenticate legitimate users with over 90% accuracy.