TBB still doesn't round windows in some cases

[cypherpunks]

I understand TBB does something to reduce the fingerprintability of the browser window size, but apparently it isn't enough as ​https://panopticlick.eff.org/ says they've never seen my screen size before. I'm using the latest TBB on the latest debian stable with the dwm window manager.

Could TBB please round the size to a larger interval so that I look like a more typical user? Thanks.

[randybytes]

Confirming this also effects TBB on windows 8.1. I think this should be a high priority bug.

[randybytes]

it looks like someone has already attempted to block the screenX and screenY, but it is not happening, window.screen property has the correct information available. ​https://github.com/tomrittervg/tor-browser/blob/70783c6cdb4bb0ce88fa17e814dc67fddc6ce078/dom/base/nsGlobalWindow.cpp#L4352

[mcs]

Replying to cypherpunks:

I understand TBB does something to reduce the fingerprintability of the browser window size, but apparently it isn't enough as ​https://panopticlick.eff.org/ says they've never seen my screen size before. I'm using the latest TBB on the latest debian stable with the dwm window manager.

Could TBB please round the size to a larger interval so that I look like a more typical user? Thanks.

For screen size, Tor Browser reports your window size. Unless you resized the browser window, the size should be a multiple of 100. What size was reported by ​https://panopticlick.eff.org/ ?

Also see #13650.

[mcs]

Replying to randybytes:

it looks like someone has already attempted to block the screenX and screenY, but it is not happening, window.screen property has the correct information available. ​https://github.com/tomrittervg/tor-browser/blob/70783c6cdb4bb0ce88fa17e814dc67fddc6ce078/dom/base/nsGlobalWindow.cpp#L4352

Access to properties within window.screen has been patched as well. Are you seeing a case where window.screen leaks the actual display dimensions or other info?

[randybytes]

Replying to mcs:

Access to properties within window.screen has been patched as well. Are you seeing a case where window.screen leaks the actual display dimensions or other info?

Are you seeing a case where window.screen leaks the actual display dimensions or other info?

Yes, on the Tor Browser bundle 4.03 with windows 8.1 leaks the actual display dimensions:

On ​https://panopticlick.eff.org it leaks:

Screen Size and Color Depth: 1366x633x24

which only 1 in 82820.68 browsers have this value.

from the javascript console window.screen shows:

Screen { availWidth: 1366, availHeight: 383, width: 1366, height: 383, colorDepth: 24, ...

Thanks for replying, is their any diagnostic information that could help?

[cypherpunks]

Replying to cypherpunks:

I understand TBB does something to reduce the fingerprintability of the browser window size, but apparently it isn't enough as ​https://panopticlick.eff.org/ says they've never seen my screen size before. I'm using the latest TBB on the latest debian stable with the dwm window manager.

Could TBB please round the size to a larger interval so that I look like a more typical user? Thanks.

Window size quantization seems to be iffy with tiling dwm. If you're using the tiling mode, try switching to stacking.

[gk]

Replying to randybytes:

Replying to mcs:

Access to properties within window.screen has been patched as well. Are you seeing a case where window.screen leaks the actual display dimensions or other info?

Are you seeing a case where window.screen leaks the actual display dimensions or other info?

Yes, on the Tor Browser bundle 4.03 with windows 8.1 leaks the actual display dimensions:

On ​https://panopticlick.eff.org it leaks:

Screen Size and Color Depth: 1366x633x24

which only 1 in 82820.68 browsers have this value.

from the javascript console window.screen shows:

Screen { availWidth: 1366, availHeight: 383, width: 1366, height: 383, colorDepth: 24, ...

Thanks for replying, is their any diagnostic information that could help?

Are you resizing/maximizing your browser window? If so, then this is the cause of the unusual screen size. Our defense is not working with resized/maximized windows yet.

[randybytes]

Replying to gk:

Replying to randybytes:

Replying to mcs:

Access to properties within window.screen has been patched as well. Are you seeing a case where window.screen leaks the actual display dimensions or other info?

Are you seeing a case where window.screen leaks the actual display dimensions or other info?

Yes, on the Tor Browser bundle 4.03 with windows 8.1 leaks the actual display dimensions:

On ​https://panopticlick.eff.org it leaks:

Screen Size and Color Depth: 1366x633x24

which only 1 in 82820.68 browsers have this value.

from the javascript console window.screen shows:

Screen { availWidth: 1366, availHeight: 383, width: 1366, height: 383, colorDepth: 24, ...

Thanks for replying, is their any diagnostic information that could help?

Are you resizing/maximizing your browser window? If so, then this is the cause of the unusual screen size. Our defense is not working with resized/maximized windows yet.

When I start the browser in windowed mode, without any resizing or maximization I get:

Screen Size and Color Depth:
one in x browsers have this value: 621890.75
value: 1004x535x24

So even with no alterations to the window, I am not getting any protection on my platform. 1 in 62K could identify my computer.

[gk]

Replying to randybytes:

When I start the browser in windowed mode, without any resizing or maximization I get:

Screen Size and Color Depth:
one in x browsers have this value: 621890.75
value: 1004x535x24

So even with no alterations to the window, I am not getting any protection on my platform. 1 in 62K could identify my computer.

Could you set extensions.torbutton.loglevel to 3 and click on New Identity? I'd like to have a look at the log output after "Torbutton INFO: New window" which you should find in your browser console (CTRL + SHIFT + J).

[mikeperry]

I am really hoping we can call this a dup of #14429, but putting this month's tag on it to track that.

[mikeperry]

Move over remaining June items to July

[joebt]

Other case where TBB (Win) still doesn't round screen size correctly (now in v6.08).

If Windows' DPI is left default 96, then as of 01/05/2017, sites report TBB initial start size 1000 W x 900 H, on a 1920x1080, 4:3, 22 inch monitor. Is that the best possible on this size / shape monitor? Since 50 - 75 or 90% ?? of desktop monitors today are probably widescreen (common 4:3), why a nearly square TBB screen that's < half the monitor?

But for many folks with < 20/20 vision or focusing issues, when Windows DPI is increased (say 110), then sites still report 1000 W, but an odd height (like 729) - never a multiple of 100.

That's starting TBB normally & not touching a button. Same issue in a fresh TBB D/L, clean install in new folder, leaving all default settings.

Why can TBB correctly round the width but not height when system DPI is changed? AFAIK, it's never correctly rounded both height & width, if Widows DPI was changed from default.

[Thorin]

Close this as dupe of #14429 ?

As for the DPI issue in comment23 one post up, this could is a dupe of #21341 ?