Re: [Help-gnunet] user feedback

[Bram Kivenko]

> Well, unlike in other systems, the popular keywords are not obvious from the
> network traffic. Suppose you see that
> A2D0761D5DC601C74E98386CDDC360FDF130B48F is a popular query
> on the network. What is the keyword? You need to produce
> 53E2DC6BE6927110E0DE04087B88C46C5F829623 in order to send a reply. If you
> figure out a way to do this in the general case, you can hijack these
> keywords; if you can not, it will be fairly difficult...

I think a dictionary attack will be very efficient.  As well, common sense
tells you that people are searching for current media topics despite the
claim that everyone is searching for "obscure, hard to get" content.

>
> > On the other hand, were a user feedback system put into place, great care
> > must be taken...since technically this is just censorship, no?
>
> Potentially, yes. A careful balance between spam detection/prevention and
> censorship must be the goal.

Spam attacks are potentially dangerous to GNU-NET.  Spam-files will
quickly replace good content to some extent.

It would be important to make it easy for the user to preview partial
content and abort transfers to limit this effect.

I also think it would be helpful in the future for a downloading client to
report to any single hop intermediaries that the aborted file was
considered spam and to flag it for early replacement -- if it decided to
put it in the repository and no one else has requested it without
reporting it as spam.  The feedback response could probably be propagated
if the intermediaries want to remember all the blocks passed through for a
given time period.  I'll assume that this may raise legal questions (such
as an imposed requirement at some time in the future.)

Bram