Fingerprinting isn’t always possible to defeat, and its not always possible to avoid making accounts (work and school accounts)
However, it should be possible to fill up tracked data with meaningless garbage and reduce the signal-to-noise ratio. Ex: a bot that browses random products on amazon to reduce profiling accuracy.
Do you guys know of any tools that do this? Anything from browser extensions to command line scripts, to anonymous group-accounts.
Having your browser lie about every detail is anonymous, but not k-anonymous. i.e. Nobody will know who you are, but your browser fingerprint is unique and so you will not blend in with everyone else. The Tor Browser and Mullvad Browser try to be k-anonymous, so everyone looks the same. Brave Browser is an interesting case where all fingerprint data is randomized, so you are not by definition k-anonymous, but you do blend in with all other Brave users in that it is all randomized in the same way for everyone.
In summary:
- Having your browser lie in a unique way is bad
- Having your browser lie in the same way as everyone else is good
- Having your browser lie in a random way like everyone else is still good
I would be very careful about saying Tor/Mullvad/Brave are anywhere near approaching k-anonymity… Tor Browser cannot even hide your real OS when queried from javascript, and there are current ways to detect all of those browsers independently.
I think one problem is that most people’s (general non-tech population) browser setups are completely bone-stock, and so by definition “random like everyone else” is likely already excluding all the stock users and placing you in a much smaller box to compare against.
I would be very careful about saying Tor/Mullvad/Brave are anywhere near approaching k-anonymity
I agree, but it’s the best we have so far. If you take some time to sit down and think about it, a lot of the problems with internet privacy can’t be fixed without a complete overhaul of our existing systems.
Tor Browser cannot even hide your real OS when queried from javascript
This is true, but the exception is Tails which lies about being Windows.
I wish I knew how tails does it so that I could make my Linux do it as well.
Edit: oh, it’s just spoofing the user agent af far as I can see. That doesn’t hide it being linux at all.
Lying about your host OS does nothing to protect against OS fingerprinting. Your OS can still he determined through the differences in how each OS renders and handles the Browser, and underlying architectural differences between browsers on each OS.
Are you saying Tails has a custom fork of TBB that spoofs the OS? Do you have a link to that patch?
Tails is an operating system. Try booting into Tails yourself and use various websites to see what I’m talking about: All of them report your operating system to be Windows, despite Tails being based on Debian.
Tails is an operating system
Yes, and it comes with Tor Browser, which normally does not spoof your OS when probed via javascript (only the user-agent), that is why I asked if you had a patch to the source code, which is what they would have to be using in order to do what you’re saying.
But as it stands, I am not able to verify your claims, as Tor Browser on Tails 6.7 is still showing the true OS via javascript queries for me:
There’s a page somewhere that, if you allow it to, opens hundreds of tabs in the background pointing to a bunch of lifestyle and commercial sites so your shadow profile ends up looking nothing like you. I will try to find it again.
And here it is: https://trackthis.link
Just FYI You would have to be using the same exact browser configuration you normally browse with, otherwise the fingerprint it uses will be different.
Yeah, cookies, account logins, and other stuff make it hard too. Ex: randomly exploring gmail emails at different times of day, but not actually marking emails as read.
Right, even the most secure/private browser cannot help opsec failures… if only one person visits the same website(s) at the same time every day, you are not anonymous. But we all must define our own threat models and apply what’s realistic for us individually.
Except for shared unique similarities. Fingerprinting designers know “not all data is good data” and will then filter out bad data and use hard to change charateristics, like hardware or software similarities, which can enable cross-browser fingerprinting.
What is a “shared unique similarity”? Sounds a lot like something that isn’t unique to me…
Unique to you, shared between your different browsers.
