🧞 Did you ever want to capture an HTML element as a video stream? But only the target element and its descendant elements would be captured. Parent and sibling elements would not. Check out proposal at github.com/WICG/proposals/is… and leave feedback!
6
14
1
78
This would be amazing. Would the consent dialog for this bypass the tab/window selector since the target is known?
1
1
One could argue that it doesn't need consent at all, if it omits cross origin iframes, since it's all information the site could get more easily by reading the DOM.
3
2
What about no-cors images?
2
4
Good point, but maybe there could also be a way to omit those.
2
1
You might need an even stronger version of COEP
1
4
Form controls and :visited are a problem too. This came up in a WebRTC-related group.
4
5
What about form controls is hidden from JS?
1
Maybe system specific native appearance as a fingerprinting vector?
1
1
Guess you could say the same about font antialiasing (although maybe some elements of that also change layout)
1
Replying to @jaffathecake @jyasskin @annevk @cramforce @quicksave2k @tomayac
I suspect that (like many fingerprinting aspects of text rendering) is already exposed through 2D canvas.

Oct 23, 2022 · 5:56 PM UTC

1
2
Replying to @davidbaron @jyasskin @annevk @cramforce @quicksave2k @tomayac
lol yeah. Silly me. Although, I wonder if the form controls are still capturable through SVG and foreign content.