.png){kind=icon}
.png){kind=icon}
Return
PassStyles: Graphical Authentication Using Mixed Face Styles
.png){kind=icon}
PRE
AI
DOI：10.1109/ACCESS.2025.3612089.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
We propose PassStyles, a graphical authentication system that utilizes images generated by StyleGAN. PassStyles authenticates users who successfully select one correct image from nine generated images multiple times in a row. The benefit of PassStyles is that the user can authenticate intuitively by choosing an image from nine images. The other benefit is its robustness to shoulder surfing because the different images generated by StyleGAN and the three-by-three combination of those images make it difficult for non-genuine users to identify the password. Moreover, PassStyles has a registration scheme that specifies which layers of StyleGAN contain the password in order to generate a wide variety of images for authentication from a few images. In addition, it uses password distilling, which removes inappropriate authentication images, to mitigate authentication errors that occur when the generated images do not adequately match the password. The evaluation experiments indicated that various image features can be passwords. In addition, we conducted shoulder surfing experiments on PassStyles and achieved a protection rate of over 90% for some passwords.
Keywords:
Machine learning
personalization
mobile interfaces
user-adaptive interaction and personalization
generative AI
Machine learning
personalization
mobile interfaces
user-adaptive interaction and personalization
generative AI
Journal
IF:
3.6
Papers:
9.8W
Citations:
29.4W