Re: My Gnutella spam reduction idea revised

Zlatin Balevsky wrote:
> * the query with the made-up word will be very heavy on the network.
> The first few test queries will not travel far, but once the nearby
> spammers are caught other test queries will become maximum-weight.

True. Perhaps test queries should only be used for the first few 
searches per session (assuming spammers of this type are quite 
unselective about which searches they respond to, so the list of 
spammers will be useful for future searches in the same session).

> * "post-processing" of results after they arrive does not help with
> dynamic querying.  Once the ultrapeers have decreased the intensity of
> the search, its too late.

I was under the impression that most spam results currently arrive out 
of band but I could well be wrong.

> * should a spammer intercept both queries, it becomes really easy to
> figure out which word is the made up one.

I'm not sure it would be easy to spot in an automated way - I quite 
often run a general query for an artist and then more specific queries 
for certain songs to get extra results, for example. It would certainly 
raise the bar by requiring the spammers to keep state.

> * If the spammers decide to answer only to single-word queries,
> they'll still do considerable damage.

Yup, good point.

Cheers,
Michael