@mattab opened this Issue on October 3rd 2018 Member

Matomo uses fingerprinting to try and match visits into a visitor.

This sometimes leads to obvious mistakes of categorising two people from two different countries as the same visitor such as here:

better fingerpriting

How to solve this issue...

  • maybe the fingerprinting could also include the geo-located country to prevent these obvious fails?

  • maybe the fingerprinting (currently very basic), could be extended to include a new additional seed that would be more unique?
@mattab commented on October 3rd 2018 Member

In https://github.com/matomo-org/matomo/issues/4321#issuecomment-51447992 @tsteur wrote:

BTW: there is a library for this which also supports canvas fingerprinting which can bring an accuracy of up to 94%: https://github.com/Valve/fingerprintjs2 / https://github.com/Valve/fingerprintjs In combination with some other techniques this accuracy could be even increased

@voarsh commented on October 17th 2018

👍 for this

@tsteur commented on October 29th 2018 Member

FYI: What needs to be taken into consideration is this:

On my machine (MBP 2013 Core i5) + Chrome 46 the default FP process takes about 80-100ms. If you use extendedJsFonts option this time will increase up to 2000ms (cold font cache).

To speed up fingerprint computation, you can exclude font detection (~ 40ms), canvas fingerprint (~ 10ms), WebGL fingerprint (~ 35 ms), and Audio fingerprint (~30 ms).

Interesting page: https://amiunique.org/fp

Powered by GitHub Issue Mirror