It was at the Securedrop website. How did I end up there ? I read something about Sequoia and encryption and then wanted to see what Securedrop entailed.
Meanwhile I’ve raised the security settings. Still, today someone in this community (?) mentioned that Tor browser does not protect the remote to check for the OS, and now this. Color me surprised.
That would be a fail of the fingerprinting protection. A properly set up TOR browser for example should not allow that detection by any means. If you know how to detect it, please report it as a critical vulnerability.
I could think of maybe some edge case behavior in webrenderer or js cavas etc., which would mainly expose info on the specific browser and underlying hardware, but that is all of course blocked of or fixed in hardened browsers.
Further, if you have a reliable method, you could sell it off to for example Netflix, who are trying to block higher resolutions for Linux browsers but are currently foiled by changing the useragent (if you have widevine set up).
What do you mean by “properly configured”
Here is a screenshot of the default Tor Browser, installed from the repos, no config changes made. As you can see, creepjs can detect that I am using Linux.
Obviously, if you disable js, then the site doesn’t work. Not sure if there are ways to detect the OS without javascript.
One common way to analyze the OS if all else fails is to look which fonts are installed. This is done by rendering thousands of divs with some text out of sight of the user. Each div with a different font. If the div width changes compared to the default, you know a font is installed. Different OS have different sets of fonts by default. Not sure if flatpak/flatseal (or other containerization methods) could protect against that. Technically you can install the exact set of Windows fonts and uninstall all Linux fonts, but I’d expect some linux app breakage and general uglyness.
An online search I did for how to completely hide the OS without breaking most websites did not result in anything except runnjng the browser in a Windows VM.
EDIT:
Per default tor has a linux useragent. And I can’t seem to change it with the useragent switcher or with about config override. So yeah… even better.
Default linux works too ofc, I didn’t know they took that route.
Most other browsers have very specific useragents, so the main pool of same useragents will be hardened browsers anyway.
Thank you for checking
edit:
https://github.com/TheTorProject/tor-messenger-build/blob/581ba7d2f5f9c22d9c9182a45c12bcf8c1f57e6e/projects/instantbird/0001-Set-Tor-Messenger-preferences.patch#L354 would indicate it should be Windows, Ill check later.
Try it with high security settings in tor, it might be something like canvas. Did you enable any permissions for the website?
Can you share your creepjs results with tor when you have some time to check it out?
As far as I know Tor browser defaults to a Windows useragent string since years. Just double-checked by visiting a website I maintain and checked its web-server log files :
My results with LibreWolf are the same : also Windows. Plain Firefox shows the correct real OS Linux as useragent string.
How comes my useragent is Linux then? I just installed it fresh trom the arch official repos for the first time to test. Creepjs shows the useragent further down (not in this screenshot) and I visited other test sites as well.
I’ll test it tomorrow by downloading it from the website.
Here some proof of my claim that Tor browser useragent string is Windows : https://wiki.archlinux.org/title/Firefox/Privacy#Change_user_agent_and_platform
From five years ago, afaik suggesting Windows is the default : https://gitlab.torproject.org/tpo/applications/tor-browser/-/issues/26146
Tor browser from the arch repos is not stock torbrowser. Add repos for torproject/guardian project/whatever it’s called now, or use the torproject.org installer.Are you saying that Arch Linux is providing Tor browser with a different useragent string ?
I believe that is the case, if you inspected the HTTP headers and found if to show Linux instead of Windows. my last experience with that would have been years ago. Arch does like to compile things from source instead of using binary blobs, and compilers and configs can undo a lot of the work the torproject has done to combat fingerprinting, which is why it’s recommended to run the pre-built binary and install no plugins. However it’s important to note that it ALSO gives you a unique JavaScript fingerprint every time, when tools use as much information as possible to generate a fingerprint, because it generates new information on every reload. That’s why OPSEC is important and for can’t help you if you use it wrong. If you login to 2 different unlinked sites in the same session, and you don’t want them to be linked, too bad now they’re linked via JS fingerprinting. JavaScript is more or less a programming language within the browser, and you’ll never escape JavaScript fingerprinting. Which is why it’s important to learn how to use tor properly, and leave JS disabled as much as you can.
One thing you can do with your arch build is use the fingerprinting tool to see how unique you are, then get a new identity, then go back and do it again. Does it now say you’re one of 2 people who have used the tool, or does it show you’re (again) unique? If the latter, then it’s working (at least enough) properly.
Interesting. I would have expected the useragent string to be part of the user configuration files that are automatically created in the browser profile directory. I’ve not used Tor browser package from Arch yet but curious to do some testing.
Yes it is. Since what the arch repos have is a script to download TB from torproject.org
You’re thinking of F-Droid (on Android), at least that is what the Guardian Project repo is for.
Thanks for the correction.
Edit. I see what you mean now. CreepJS indeed shows userAgent Linux. Weird. It got the amount of cores wrong btw.
Becuase creepjs uses JS to figure out the userAgent, WorkerNavigator.userAgent
It isn’t.
Yes it should, through Javascript. https://gitlab.torproject.org/tpo/applications/tor-browser/-/issues/26146#note_2649490
Blocking UA access via JS alone is not enough.
Do you have a better source than a 5 y/o comment in an issue?
Trust me, I spent 3 years in the Tor community I know this shit, this thing comes up so often.
Alternatively: https://gitlab.torproject.org/tpo/applications/tor-browser/-/issues/41610