Server-side commercial fingerprinting

https://dl.acm.org/doi/10.1145/3774904.3792687 - open access

Researchers subscribed to Fingerprint Pro and performed tests on the algorithm using FP-Spoofer, which involved systematically varying the visible attributes of the browser. They tried to find out what variations were needed to make FP produce a new identifier for a browser between sessions.

Interestingly, Brave’s protections failed against its server-side fingerprinting , while Firefox with RFP (but not FPP) achieved cross-session unlinkability. Brave required a VPN on top of its farbling approach to defeat it, while Firefox with RFP did not.

Using a VPN increased the chances that a variation would produce a new identifier, while private browsing modes did not. The algorithm apparently allows for more variation before issuing a new identifier when private browsing mode is recognised. Brave apparently suffered from a similar dynamic as private browsing modes, as Brave browsers were internally labelled effectively as aiming to defend against the algorithm. It’s unclear why Tor or Firefox with RFP did not result in similar labelling.

People talk say stuff like, “this makes you more fingerprintable”, “Changing this makes you more fingerprintable”. What kind of fingerprinting are we talking about when we say stuff like this? Simple fingerprinting scripts? Fingerprinting as a service (FASS) like Fingerprint Pro? Sysadmins looking at logs?

I suppose it depends on your threat model. For surveillance capitalism by big tech I would assume it’s FASS, in which case you’re either detected as the same user or not, although I do wonder if in a law enforcement investigation they would be looking through the raw data and the putting more of a human into the automated tool, kinda like how Monero tracing with eliminating ring signatures works.

It’s so insidious they try to essentially give everyone an account without signing up. “Identify Every Visitor… even if they’re anonymous.”

1 Like

Usually I think it’s saying, you’re changing this setting from the default and that can be detected. Like if you set your browser to request websites follow your system theme and it’s dark, websites can detect that.

I did a handful of tests on the fingerprint pro demo.

For Firefox and Safari (with fingerprinting mitigations on), merely changing the connection to my mobile hotspot was sufficient to register a new identifier.
For Brave, alternatively, connecting to a VPN in a different country while additionally spoofing the time zone to match it was the minimum required that I found that achieved the same.

Fingerprint pro doesn’t like Brave. Brave for its part claims this kind of result is a false positive caused by trying multiple times close together, but this doesn’t explain the difference with Firefox and Safari, where less change is needed to produce a new identifier.

Its weird that Brave advertises itself as Brave (as if it is advertising) in Client Hints and navigator.brave.isBrave(), which likely makes this strategy by Fingerprint Pro easier:

“In general, it is not a privacy goal for the Brave browser to hide from websites that the user is using Brave.”