Hi there,
Stumbled across the site and enjoyed reading the papers. :)
I’m curious: Have you looked into whether the entropy distribution varies for different traffic types? For instance, since the measurement is based on IP, the use of mobile carrier proxies might impact the entropy calculation. Mobile traffic may look different from desktop in this regard.
Let me know if i should redirect questions to another channel. Thanks!
Hi @swuen,
Thanks for your interest in the project and sharing feedback with us.
Your comment is very appropriate. It is true that Proxies or NATs which aggregate a large number of users behind a single IP might be an issue in the computation of entropy. There are ways to address this issue, for instance using other user identifiers instead of the IP address such as the pair <IP address, User Agent> or even a cookie id. Using these ids will solve the problem of multiple users assigned to a single identifier (i.e., the IP address). However, these ids have their own problems. For instance, it is very hard for an attacker to fake an IP address whereas faking/modifying a user agent or faking/modifying/eliminating a cookie id is trivial.
This is an exciting debate. If you have ideas about it, we are happy to listen to them. But probably it is better to discuss it privately to avoid creating too much noise.
Hi guys,
I followed the quide of installing 'nameless' at https://github.com/Nameles-Org/Nameles .
But how can I begin to use it? Do you have the practic docs of how to use it?
I'am sorry for the silly question.
Thanks.
As the section 3 of the readme says, we do not have yet an API that you can use as a library. But, you can use the dsp_latency_test.cpp code in the dsp-emulator as a template of how to write methods in your program for sending queries to the scoring module and how to handle the replies.
The test code is in c++, but you can implement a similar scheme in any language for which zeromq is available.
Thanks for stopping by and write me if you have any specific question.
