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

--- In [email protected], "Sam Berlin" <sberlin@...> wrote:
> FWIW, I have no personal opposition to injecting 'ad-style' results
> into the network, so long as there's a mechanism for displaying them
> as something other than spam-search-results.  Perhaps we should
> consider thinking of ways to allow these business-minded folks to
> display these ads in a friendly way.  Then we can let the user
> decide if they want to see relevent ads, and concentrate better 
> on getting rid of the real spam.

   Makes sense. To be clear, when I was talking about spam, I was
thinking about such things as the nodes that answer any query with
something like "<your_query_here>_network_accelerator.exe"; if the
reply is relevant to the query, then it is more or less a search
result rather than a spam. As long as I'm not getting Viagra
commercials while looking for music, these are indeed just the
sponsored results, and the only issues are just the fair division 
of the screen real estate between these and the normal search
results, and maybe - as you say - the clear marking of the ads.

   So I'd say that we have to distinguish between sponsored content
(paid for, but relevant files), and spam - the results that are not
relevant to the search terms. That would be the links that push
Viagra regardless of what you were asking for, to give just one
example.

   And these two types of content should be treated differently. 
In fact, I'm not even sure if the sponsored results should indeed 
be always marked as such. For example, if someone like Skyrider
injects the ads into the Harry Potter movie that I've been asking
for, it is no different from the normal Harry Potter movie grabbed
from TV - so why should we display such a movie in a sponsored
content ghetto? It would be nice to know about the ads in advance,
but this is a problem for someone like Bitzi, not the spam issue
per se. Probably makes sense to mark such a movie as coming from
Skyrider, but the Skyrider folks might be happy to do so in any 
case (a good indicator of a difference between the spam and the
legit ad might be that the legit ad is not afraid to be clearly
seen as such, whereas spam usually has to trick the user into
even opening it).

   Of course, I'm sure that the issue is not that simple, and for
most search terms there can be a whole spectrum of relevance (i.e.
Harry Potter movie, sponsored by Hollywood to be promoted on the
Gnutella network -> Harry Potter movie trailer -> Harry Potter
poster -> Harry Potter store on Amazon -> "Harry the Potter - 
best pots in Oklahoma" -> Harry_Potter_enlarge_yourself.htm).

   The way I see it, the only problem with sponsored content (for
example, the case where the musician has actually paid someone to 
make this music available on the net) is that it screws up the
rating system that is implicit in a P2P network. When you're using
a P2P net, you actually rely on things like content availability to
judge the content value and popularity - when you see something with
1000 copies, it is like a high Google page rank.

   Google clearly marks the sponsored links, because their page 
rank is nowhere close to their visibility on the results page, 
and it would be misleading to imply that these are the popular
pages. Ditto for Gnutella. Showing multiple copies of a sponsored
data item only because you have a large server farm seeded with it,
is misleading and hurts the network usability, because the users 
can no longer trust the network popularity metrics.

   Thus maybe the sponsored links need to be clearly marked as such,
lest the network usability as a whole would take a hit. Of course,
Google controls its web page, and Gnutella doesn't really control
what goes to the search results, so there should be some voluntary
agreement about this, and the participants not following it should
be moved into the "spam node" category and eliminated by technical
means (PeerGuardian and such).

   That's how this all looks to me, anyway.

   Best wishes -
   Oso.
   27 Mar 2007.


--- In [email protected], "Sam Berlin" <sberlin@...> wrote:
>
> I'm sure there's bots out there that are manipulating preinstalled
> LimeWires (and that's a problem), but that's not actually what I was
> referring to.  There's some degenerate code in LimeWire that allows
> hosts to easily flood their addresses into the pool of addresses and
> cause each LimeWire to try connecting to them over and over.  This
> leads to a huge portion of clients foolishly connecting to the same
> few hosts, making the network lopsided towards those hosts.
> 
> We're first looking to fix the configuration of the network.  Once
> that settles down again, we have some ideas (very hazy ones, right
> now) for spam.
> 
> FWIW, I have no personal opposition to injecting 'ad-style' results
> into the network, so long as there's a mechanism for displaying them
> as something other than spam-search-results.  Perhaps we should
> consider thinking of ways to allow these business-minded folks to
> display these ads in a friendly way.  Then we can let the user 
decide
> if they want to see relevent ads, and concentrate better on getting
> rid of the real spam.
> 
> Sam
> 
> On 3/27/07, Serguei Osokine <osokin@...> wrote:
> > --- In [email protected], "Sam Berlin" <sberlin@> wrote:
> > >
> > > There's a large range of IP addresses out there that have 
managed to
> > > insert themselves into the core of the network, by abusing 
problems
> > > within LimeWire's code.
> >
> >   Oh. So you mean that by your estimate most of the spam is caused
> > by the bots that have remotely gained control over LimeWire 
servents?
> > I didn't realize that. Sorry to hear about it.
> >
> >   But still - I dunno... Even when you fix that, I have a feeling
> > that spam is not going away any time soon. Even a perfectly legit
> > modified client can inject quite a lot of it - and since people 
have
> > actually started writing business plans that rely on polluting the
> > network with crap, getting rid of it might be a bit more complex
> > than just "toughening up the code" of one single vendor.
> >
> >   But - I wish you luck anyway.
> >
> >   Best wishes -
> >   Oso.
> >   27 Mar 2007.
> >
> >
> > --- In [email protected], "Sam Berlin" <sberlin@> wrote:
> > >
> > > There's a large range of IP addresses out there that have 
managed to
> > > insert themselves into the core of the network, by abusing 
problems
> > > within LimeWire's code.  These problems range from being not 
being
> > > wary enough of network-damaging messages to plain old bad code 
(such
> > > as thinking the first item [hashcode-wise] in a hash map is 
random).
> > >
> > > We've got some good ideas planned for toughening up the code, 
which
> > > should make damaging the network much harder.  This isn't 
addressing
> > > spam as a "spam is the problem" approach, but should have much 
of
> > the
> > > same effect.
> > >
> > > Sam
> > >
> > > On 3/27/07, Serguei Osokine <osokin@> wrote:
> > > > On Sunday, March 25, 2007 Zlatin Balevsky wrote:
> > > > > Any protocol-level ideas will be published here.  So far we
> > have the
> > > > > OOBv3 proposal.
> > > > >
> > > > > Any limewire-specific measures will be discussed in 
limewire-
> > specific
> > > > > places.
> > > >
> > > >        I was not talking about measures. I was talking about 
the
> > data
> > > > that I thought you might already have, assuming that you spent
> > some
> > > > time looking into spam-related issues. Is the spam coming from
> > some
> > > > relatively limited IP ranges, or it is injected by the routers
> > into
> > > > the spoofed TCP sessions that make it look like coming from 
the
> > > > legitimate servent IPs; is it possible to filter out such IP
> > ranges,
> > > > assuming that they exist; does PeerGuardian already know about
> > these
> > > > ranges or it is not tracking spam; does the situation improve 
with
> > > > just display blocking of the results, or the spam actually 
maxes
> > out
> > > > some result number limits en route, and filtering it out 
leaves
> > you
> > > > with few useful results; do the spam sources migrate from one 
IP
> > to
> > > > another a lot - this kind of thing.
> > > >
> > > >        If you don't have this data - just say so. If you have 
it,
> > but
> > > > do not want to discuss it - that's cool, too. All this can be
> > collected
> > > > independently; it will just take some time.
> > > >
> > > >        Best wishes -
> > > >        Oso.
> > > >        26 Mar 2007.
> > > >
> > > >
> > > > -----Original Message-----
> > > > From: [email protected] [mailto:[email protected]]
On
> > Behalf
> > > > Of Zlatin Balevsky
> > > > Sent: Sunday, March 25, 2007 9:36 PM
> > > > To: [email protected]
> > > > Subject: Re: [the_gdf] Drop Dynamic Querying - It only 
benefits
> > Spammers
> > > > these days
> > > >
> > > >
> > > > Any protocol-level ideas will be published here.  So far we 
have
> > the OOBv3
> > > > proposal.
> > > >
> > > > Any limewire-specific measures will be discussed in limewire-
> > specific
> > > > places. ( http://limewire.org )
> > > >
> > > > On 24 Mar 2007 13:52:50 -0700, Serguei Osokine <osokin@> 
wrote:
> > > > >
> > > > >   On Monday, January 23, 2007 Arne Babenhauserheide wrote:
> > > > > > Could you shed some light on your thoughts?
> > > > > >
> > > > > > They seem somehow blackboxed just now :)
> > > > >
> > > > > Yeah, I'd be very interested to hear these thoughts too, 
Sam :-)
> > > > >
> > > > > Or maybe not. Maybe they are secret thoughts. Anyway, 
forgive
> > > > > me for jumping into this discussion two months late, but 
that
> > is not
> > > > > because I was not paying attention, but rather because I was
> > thinking.
> > > > > And the result of all this thinking is that the spam fight 
is
> > hopeless
> > > > > without the PeerGuardian plugin - maybe together with some 
kind
> > of a
> > > > > content filter.
> > > > >
> > > > > Let's backtrack a little. For many years, spam was not a 
serious
> > > > > problem, because it would not replicate. Who'd store it on 
his
> > machine?
> > > > > So the spam content was always present, but was not much of 
a
> > problem,
> > > > > because it comprised just a small share of all content in 
the
> > network.
> > > > >
> > > > > The situation changed when people started to deploy the
> > > > > specialized high-performance servers to inject the massive
> > amounts
> > > > > of fake content into the system. It probably started with 
the
> > various
> > > > > attempts to destroy the P2P nets with pollution, but now it 
is
> > more of
> > > > > a commercial endeavour, with people like P2Pads and Skyrider
> > using
> > > > > high-performance methods to deliver lots and lots of spam. 
(And
> > even
> > > > > claiming that they've got a novel business model. Yeah, 
right.
> > Ever
> > > > > got an anatomy enlargement e-mail offer?)
> > > > >
> > > > > So the spam comes in two forms: from their machines and 
through
> > > > > poor souls who were duped into downloading this crap and now
> > are the
> > > > > rebroadcasters. I'd say that the first channel is more 
serious
> > (that
> > > > > is, it produces lots and lots of new spamming content in
> > virtually
> > > > > unlimited quantities). Fortunately, it is also easier to 
break:
> > it
> > > > > is hard to constantly move a massive server farm from one IP
> > range
> > > > > to another, so I think that the PeerGuardian plugin can take
> > care
> > > > > of it.
> > > > >
> > > > > The second part (killing the spam that is already present) 
is
> > > > > trickier, but once the inflow of this crap is limited, maybe
> > Credence
> > > > > or something like that can help (right now it must be 
basically
> > useless,
> > > > > because the spamming farm can mutate its content all the 
time).
> > Heck,
> > > > > maybe with the spam farms blocked by the PeerGuardian from 
the
> > majority
> > > > > of the network nodes, the usual natural selection process 
would
> > be
> > > > > enough to bring the spamming problem into reasonable bounds.
> > After
> > > > > all, any node that sees a result from the blocked IP range 
can
> > kill
> > > > > it, so hopefully much of the spam can be blocked by a small
> > percentage
> > > > > of nodes having such a plugin. (Though direct replies do
> > complicate
> > > > > things, of course.) Maybe having a PeerGuardian plugin 
should
> > be a
> > > > > requirement for at least every Ultrapeer?
> > > > >
> > > > > Anyway, the rebroadcasters lack one very important quality:
> > > > > they can no longer create the artificial spam file name and
> > stuff
> > > > > it as a result into every request coming through them. I'd 
guess
> > > > > that most of the spam is generated by the spam farms, and 
taking
> > > > > them out should lower the spam volume a lot. We should do it
> > first,
> > > > > see what happens, and go from there.
> > > > >
> > > > > And the beauty of this situation is that this spamming 
problem
> > > > > is capable of giving a huge boost to the PeerGuardian
> > development -
> > > > > from a privacy protection service (which is not something 
that
> > many
> > > > > people are worried about), it becomes a spam-fighting 
service,
> > which
> > > > > is something that everyone is aware of. Everyone sees spam, 
and
> > is
> > > > > totally sick and tired of it. With a few donations from the
> > servent
> > > > > vendors and from the community, PeerGuardian might really
> > become a
> > > > > major spam-fighting service and a household word.
> > > > >
> > > > > Of course, this integration with PeerGuardian requires not
> > > > > only blocking the connections to the blacklisted IPs (as it 
is
> > > > > normally done), but it should also include the filtering of 
the
> > > > > result list on every node for these same IPs. Every node 
should
> > > > > remove the spam results once it sees them. And since it also
> > > > > includes the displaying (requestor) node, such an 
integration
> > > > > should bring immediate benefits to the node user - he should
> > stop
> > > > > seeing many spam results at once. So the upgrade to the 
servent
> > > > > version that takes such a plugin should be a snap - the 
users
> > will
> > > > > stand in line for it.
> > > > >
> > > > > The PeerGuardian scalability should be carefully controlled,
> > > > > of course - but that is not anything unsurmountable.
> > > > >
> > > > > If I'd work for P2P spammers, my next step would be to move 
my
> > > > > operation to the individual users' machines and to the 
machines
> > that
> > > > > are controlled by the bots, but I dunno... First, these
> > machines are
> > > > > as easily to identify by the PeerGuardian, and second, the 
P2P
> > spam
> > > > > operators are currently trying to position themselves as
> > legitimate
> > > > > businessmen. I'm not sure how willing they would be to 
shift to
> > the
> > > > > networks of bots controlled by the Trojans from Nigeria-
based
> > control
> > > > > center. They are carefully avoiding even the very 
word 'spam'
> > in their
> > > > > press releases lest they be confused with universally hated 
e-
> > mail
> > > > > spammers. Utilizing the same shady methods as Viagra 
pushers do
> > might
> > > > > not look good for the public and for investors.
> > > > >
> > > > > Sam, did you guys try integrating PeerGuardian or some 
analog
> > > > > into the system? What was the result? To put it shortly, 
why do
> > P2P
> > > > > nets still have spam? What is the hole in my reasoning? Or
> > maybe you
> > > > > are already shipping this thing, and I simply missed it? In 
this
> > > > > case, I want a premium version of it, right now. And never 
mind
> > my
> > > > > Java allergy - a spamless servent is worth waiting twenty
> > seconds
> > > > > for its launch :-)
> > > > >
> > > > > Best wishes -
> > > > > Oso.
> > > > > 24 Mar 2007.
> > > > >
> > > > >
> > > > > -----Original Message-----
> > > > > From: [email protected] <the_gdf%40yahoogroups.com>
> > [mailto:
> > > > > [email protected] <the_gdf%40yahoogroups.com>]On 
Behalf
> > > > > Of Arne Babenhauserheide
> > > > > Sent: Monday, January 23, 2007 11:25 PM
> > > > > To: [email protected] <the_gdf%40yahoogroups.com>
> > > > > Subject: Re: [the_gdf] Drop Dynamic Querying - It only 
benefits
> > Spammers
> > > > > these days
> > > > >
> > > > > Could you shed some light on your thoughts?
> > > > >
> > > > > They seem somehow blackboxed just now :)
> > > > >
> > > > > I know that LimeWire is the biggest player in the GNet at 
the
> > moment,
> > > > > but
> > > > > LimeWire can only reach a certain kind of users, so there 
are
> > quite a
> > > > > lot of
> > > > > niches for other programs, and I for my part want to see
> > Gnutella a
> > > > > vivid
> > > > > community which evolves ever further into the most 
efficient p2p
> > > > > network.
> > > > >
> > > > > And I believe, that this can only happen, if different
> > developers
> > > > > with a lot
> > > > > of different ideas and ways of thinking work together.
> > > > >
> > > > > And aside from that, a heterogenous network is far harder to
> > > > > attack :)
> > > > >
> > > > > So: What's your thoughts till now?
> > > > >
> > > > > Besides: Why don't you use the description of Gnutella from
> > > > > http://gnufu.net
> > > > > on the LimeWire pages? They are intended to be used :)
> > > > > If it's licensing problems, you just need to ask me to get a
> > > > > differently
> > > > > licensed version (of the parts I wrote, which is most of the
> > texts).
> > > > >
> > > > > Best wishes,
> > > > > Arne
> > > > >
> > > > > El Martes, 23 de Enero de 2007 21:11, Sam Berlin escribi�:
> > > > > > We are actively looking into the spam problem, Arne, and
> > intend to
> > > > > > address it effectively.
> > > > > >
> > > > > > Sam
> > > > >
> > > > > [Non-text portions of this message have been removed]
> > > > >
> > > > > --- End forwarded message ---
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > > [Non-text portions of this message have been removed]
> > > >
> > > >
> > > >
> > > >
> > > > Yahoo! Groups Links
> > > >
> > > >
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > Yahoo! Groups Links
> > > >
> > > >
> > > >
> > > >
> > >
> >
> >
> >
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
>