Or, we could just look at the keyboard as they type and gain a lot more information.
In an absolute sense not showing anything is safer. But it never really matters and just acts as a paper cut for all.
> When trained on keystrokes recorded by a nearby phone, the classifier achieved an accuracy of 95%, the highest accuracy seen without the use of a language model. When trained on keystrokes recorded using the video-conferencing software Zoom, an accuracy of 93% was achieved, a new best for the medium.
Have you ever watched a fast touch typist, someone that does over 100 words per minute? Someone who might be using an keyboard layout that you're not familiar with? When the full password is entered in less than a second it can be very difficult to discern what they typed unless you're actually recording with video.
But sure, if you're watching someone who types with one finger. Yes, I can see that.
Besides, observe that several times and you might get close. Look at the stars several times and learn nothing beyond what you learned the first time.
This whole type of attack hinges on the user using weak passwords with predictable elements in any case.