Re: [Bitcoin-development] Bloom bait

[Peter Todd <pete@petertodd.org>]

[-- Attachment #1: Type: text/plain, Size: 2347 bytes --]

On Sat, Jun 07, 2014 at 03:44:07PM -0400, Alan Reiner wrote:
> 
> On 06/07/2014 07:22 AM, Mike Hearn wrote:
> >
> > You can send different bloom filters to different peers too, so I'm
> > not sure why you're listing subsetting as a unique advantage of prefix
> > filters.
> >
> >
> 
> Please let me know if we've gone down this path before, but it would
> seem that the more different bloom filters you create, the more
> information you give away.  It would be most useful to create a single
> bloom filter that captures every address you ever intend to use (say a
> look ahead of 1000 addresses), and then only ever communicate that. 
> Once people see multiple filters that you produce, they can start
> looking at the intersection of them to reduce the identity space.  I
> would expect that after enough bloom variants, they could figure out a
> perfect subset of blockchain addresses in your wallet.  (I suppose you
> could intentionally select an extra 20% addresses to include in every
> bloom filter, but it's a hack).
> 
> Similarly, if you keep updating your bloom filter to include more
> addresses, the difference in what passes through the previous one and
> the new one gives away information about new addresses you created.

You're completely correct. You can use the same nTweak value for each
filter and then slice up the filters bitwise, but then you end up
linking every query you make to your identity unless you just used a
fixed nTweak that everyone else uses.  (e.g. zero) If you do that, you
still have the problem that you're greatly increasing the load on the
network.

In any case, all this shows is that in the future bloom filters will
very likely go away eventually, and to make that upgrade path smooth it
only makes sense to define a way for nodes to let others know whether or
not bloom is supported. A NODE_BLOOM service bit is a very reasonable
and simple way to do exactly that, and is defacto what implementations
that don't support bloom filters do anyway.

Note BTW that re: DNS seeds, once the NODE_BLOOM BIP is accepted and the
NODE_BLOOM patch merged into bitcoind, I'll write a patch for sipa's
seeder to make it only return seeds with bloom filter support.

-- 
'peter'[:-1]@petertodd.org
00000000000000003afb1fdf0867fc063775e69f9ae79870bb8727f25b49e88f

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 685 bytes --]