Re: Drop Dynamic Querying - It only benefits Spammers these days

El Sábado, 27 de Enero de 2007 21:54, Philippe Verdy escribió:
> The solution is then not to change the diameter of searches, but to make
> search routing even more specific. This is where a DHT-based routing will
> help (but it will not avoid spammers).

I don't think that DHT based routing will be a solution for keyword based 
searches, because DHTs are most efficient when you search for a hash and 
Gnutella is vastly efficient for keyword searches. 

It will be very useful, however, to get more sources for a stale download or a 
magnet-link. 

> Another interesting idea is to stop sending results through the Gnet, and
> use direct routing for replies (OOB replies, through TCP or UDP), so that
> results can no more be intercepted so easily. 

Are results really intercepted? 
There is only a maximum of 4 hops between the replier and the querier, and if 
the spammers are that wide spread, that they can efficiently invade there, 
then we simply have lost. 

Instead it might be possible to use the intermediary hosts as spam-filters. 

When a user marks a file as spam, the client could simply stop the query reply 
(that's what GTKG already implements, but using a static spam-list instead of 
a spam list which the users can add to). 

> But for effectively fighting spam, we need a parallel strategy to allow
> detecting and monitoring the level of pollutions. This is not easy, because
> even spammers could also send faked bad reputation notices to increase the
> presence of their own replies, as they will also usethis system to increase
> their reputation level with faked "good reputation notices".

There was Credence to take care of that, but it seems to have stopped (being a 
university project which seems to be done now). The idea looked very good, 
though. I assume it could still have been done much more efficient, using 
more of the Gnutella infrastructure, I think. 

http://credence-p2p.org

> So finally, the only thing that will limit spamis to have more non-spamming
> nodes in the network than spamming nodes. This is exactly the same problem
> with spam through email or IM, where there's no definitive solution other
> than having a good number of wellknown trustable sources of information
> about the presence of spammers, and an active collaborating community.
>
> Running a trustable reputation system can become very costly, and this is
> where the whole business of antispam-system providers are working to help
> their customers. But is there someone ready to pay for this service, or
> readyto support this huge charge for free?

Definitely not (at least not for us...), so we need a decentral solution, I 
think. 

Best wishes, 
Arne


[Non-text portions of this message have been removed]