From: Peter Todd <pete@petertodd•org>
To: Gregory Maxwell <gmaxwell@gmail•com>
Cc: Bitcoin Development <bitcoin-development@lists•sourceforge.net>
Subject: Re: [Bitcoin-development] Bait for reusable addresses
Date: Fri, 24 Jan 2014 04:02:18 -0500 [thread overview]
Message-ID: <20140124090218.GA15398@savin> (raw)
In-Reply-To: <CAAS2fgQmsxjkQFSiCdeMoVMaqq5720KpUpdkKZOE+XytHsWw0w@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 4195 bytes --]
On Wed, Jan 15, 2014 at 05:23:04PM -0800, Gregory Maxwell wrote:
> It also has a downside of not being indexable for the server, the
> server must do O(clients * reusable-address-txn) work and the work
> includes an ECC multiply.
>
> An idea that Adam Back had originally proposed was including optional
> "bloom bait", a small token— say 8 bits— that distinguished
> transactions which allowed an anonymity set vs filtering trade off.
> Such a bait would be indexable, enabling faster lookup too.
>
> But bloom bait has privacy problems more severe than the current SPV
> bloom filtering. While you leak information to your SPV servers today
> if you use bloom filtering the leak usually goes no further. So a
> compromise requires both a statistical attack _and_ using SPV servers
> that log data against your interest. With bloom bait the whole
> network can see the relation. That is unfortunate.
Yes, but remember I proposed prefixing in my blockchain data query paper
because it's a trade-off between theoretical good privacy and
brittleness. The real world experience is that users, or to be exact
wallet authors, turn down SPV privacy parameters until bloom filters
have almost no privacy in exchange for little bandwidth usage. (though
load on the server is unchanged of course)
The brittleness comes in because the moment you connect to a malicious,
data-collecting peer, the contents of your wallet are all revealed.
Frankly that'd be a disaster for CoinJoin too, and I think it'd be a
bigger disaster than the poor specificity patterns leaked by prefix
usage. If anyone wants to deanonymize CoinJoin there will be a lot of
incentives to do so, and you only need wallet content data to do that.
> I suggest instead that with optional bait is included in an address
> that the sender compute H(nonce-pubkey) and then pick one byte at
> random out of the first 16 and xor it with the specified bait and
> store the result in the transaction. An SPV server can now index the
> bait as it comes in by extracting 16 8-bit keys from each transaction
> (the 16 bytes xored with the bait in the transaction). When the
> client wants to search for transactions it can give the server a list
> of keys its interested in— including their real key and number of
> random number of cover keys.
>
> I didn't give any though into the parameters 8-bits and 16 dimensions.
> Some reasoning should be done to fix the parameters in order to make
> them the most useful: e.g.
>
> Systems derived from more complex linear codes might give better
> performance, e.g. two secret bloom baits, two prefixes in the
> transaction bait0^random_char[0-8], bait1^random_char[0-8], server
> extracts 16 keys.. and returns to the client transactions which have
> at least two key matches with their list.
>
> Obviously whatever is used needs to be easy to implement, but schemes
> loosely based on fountain codes should only require picking some
> things and xoring... so they should be simple enough.
Well, that's the big question: How much extra data do we need and what's
the chance that this will get turned into miner-committed indexes? Or
even just provided at all? We keep on saying that miner-commitments may
next happen at all because of performance issues, and adding n extra
indexes doesn't exactly help that situation. I really suspect that the
moment that gets implemented we'll see wallet software use that for
simple security reasons, so plan ahead for that.
In the short term without miner-commitments it's just a question of how
much extra load we subject servers to. Again, getting people to even
implement prefixes isn't a trivial argument to make, yet bloom has some
serious scalability problems. (though does do roughly what you're
proposing)
In any case, your "bait" proposal is stealth address specific - how
would you propose applying the same principle to all addresses? Again,
it's a tradeoff between brittleness - connecting to a malicious peer
reveals your wallet - and blockchain stats data.
--
'peter'[:-1]@petertodd.org
0000000000000001315c71472fdce344f85f794a7135e25554f2b51dfa6b83c4
[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 685 bytes --]
next prev parent reply other threads:[~2014-01-24 9:02 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-01-16 1:23 Gregory Maxwell
2014-01-24 9:02 ` Peter Todd [this message]
2014-01-24 12:26 ` Mike Hearn
2014-01-24 15:26 ` Peter Todd
2014-01-24 21:58 ` Jeremy Spilman
2014-01-24 23:15 ` Mike Hearn
2014-01-24 15:42 ` Adam Back
2014-01-24 16:13 ` Peter Todd
2014-01-25 16:19 ` [Bitcoin-development] (space) efficient reusable addr via weil pairing IBE (Re: Bait for reusable addresses) Adam Back
2014-02-02 2:36 ` Peter Todd
2014-02-02 9:16 ` Jeremy Spilman
2014-02-02 12:26 ` Peter Todd
2014-02-02 15:26 ` Adam Back
2014-02-02 11:55 ` Peter Todd
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140124090218.GA15398@savin \
--to=pete@petertodd$(echo .)org \
--cc=bitcoin-development@lists$(echo .)sourceforge.net \
--cc=gmaxwell@gmail$(echo .)com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox