Re: [bitcoin-dev] Committed bloom filters for improved wallet performance and SPV security

From: bfd@cock•lu
To: Aaron Voisine <voisine@gmail•com>
Cc: Bitcoin Protocol Discussion <bitcoin-dev@lists•linuxfoundation.org>
Subject: Re: [bitcoin-dev] Committed bloom filters for improved wallet performance and SPV security
Date: Tue, 03 Jan 2017 16:10:14 -0800	[thread overview]
Message-ID: <22b7d05fb2b8a7a0f1c2fa0b6b375f7e@cock.lu> (raw)
In-Reply-To: <CACq0ZD5WV7ORmEJgGSquyRzndH_XP9FrLbwSNKQC5Zuh08NVDw@mail.gmail.com>

Unfortunately a non validating SPV wallet has absolutely no idea if
the information about an unconfirmed transaction they are seeing is
anything but properly formatted. They are connecting to an easily
manipulated, sybil attacked, and untrusted network and then asking
them for financial information. Seeing an unconfirmed transaction in a
wallet that's not also fully validating is at best meaningless.

On 2017-01-03 15:46, Aaron Voisine wrote:
> If the sender doesn't control the receiver's network connection, then
> the information the receiver gains by watching the mempool is if the
> transaction has propagated across the bitcoin network. This is useful
> to know in all kinds of situations.
> 
> Aaron Voisine
> co-founder and CEO
> breadwallet [2]
> On Tue, Jan 3, 2017 at 3:06 PM, adiabat <rx@awsomnet•org> wrote:
> 
>> Mempool transactions have their place, but "unconfirmed" and "SPV"
>> don't belong together.  Only a full node can tell if a transaction
>> may get confirmed, or is nonsense.  Unfortunately all the light /
>> SPV wallets I know of show mempool transactions, which makes it hard
>> to go back... (e.g. "why doesn't your software show 0-conf! your
>> wallet is broken!", somewhat akin to people complaining about RBF)
>> 
>> So, this is easy, just don't worry about mempool filtering.  Why are
>> light clients looking at the mempool anyway?  Maybe if there were
>> some way to provide SPV proofs of all inputs, but that's a bit of a
>> mess for full nodes to do.
>> 
>> Without mempool filtering, I think the committed bloom filters would
>> be a great improvement over the current bloom filter setup,
>> especially for lightning network use cases (with lightning, not
>> finding out about a transaction can make you lose money).  I want to
>> work on it and may be able to at some point as it's somewhat related
>> to lightning.
>> 
>> Also, if you're running a light client, and storing the filters the
>> way you store block headers, there's really no reason to go all the
>> way back to height 0.  You can start grabbing headers at some point
>> a while ago, before your set of keys was generated.  I think it'd be
>> very worth it even with GB-scale disk usage.
>> 
>> -Tadge
>> 
>> On Tue, Jan 3, 2017 at 5:18 PM, Aaron Voisine via bitcoin-dev
>> <bitcoin-dev@lists•linuxfoundation.org> wrote:
>> 
>> Unconfirmed transactions are incredibly important for real world
>> use. Merchants for instance are willing to accept credit card
>> payments of thousands of dollars and ship the goods despite the fact
>> that the transaction can be reversed up to 60 days later. There is a
>> very large cost to losing the ability to have instant transactions
>> in many or even most situations. This cost is typically well above
>> the fraud risk.
>> 
>> It's important to recognize that bitcoin serves a wide variety of
>> use cases with different profiles for time sensitivity and fraud
>> risk.
>> 
>> Aaron
>> 
>> On Tue, Jan 3, 2017 at 12:41 PM bfd--- via bitcoin-dev
>> <bitcoin-dev@lists•linuxfoundation.org> wrote:
>> The concept combined with the weak blocks system where miners commit
>> 
>> to potential transaction inclusion with fractional difficulty blocks
>> 
>> is possible. I'm not personally convinced that unconfirmed
>> transaction
>> 
>> display in a wallet is worth the privacy trade-off. The user has
>> very
>> 
>> little to gain from this knowledge until the txn is in a block.
>> 
>> On 2017-01-01 13:01, Jonas Schnelli via bitcoin-dev wrote:
>> 
>>> Hi
>> 
>>>> We introduce several concepts that rework the lightweight Bitcoin
>> 
>>>> client model in a manner which is secure, efficient and privacy
>> 
>>>> compatible.
>> 
>>>> 
>> 
>>>> The BFD can be used verbatim in replacement of BIP37, where the
>> filter
>> 
>>>> can be cached between clients without needing to be recomputed.
>> It can
>> 
>>>> also be used by normal pruned nodes to do re-scans locally of
>> their
>> 
>>>> wallet without needing to have the block data available to scan,
>> or
>> 
>>>> without reading the entire block chain from disk.
>> 
>>> I started exploring the potential of BFD after this specification.
>> 
>>> 
>> 
>>> What would be the preferred/recommended way to handle
>> 0-conf/mempool
>> 
>>> filtering – if & once BDF would have been deployed (any type,
>> 
>>> semi-trusted oracles or protocol-level/softfork)?
>> 
>>> 
>> 
>>> From the user-experience perspective, this is probably pretty
>> important
>> 
>>> (otherwise the experience will be that incoming funds can take
>> serval
>> 
>>> minutes to hours until they appear).
>> 
>>> Using BIP37 bloom filters just for mempool filtering would
>> obviously
>> 
>>> result in the same unwanted privacy-setup.
>> 
>>> 
>> 
>>> </jonas>
>> 
>>> 
>> 
>>> 
>> 
>>> _______________________________________________
>> 
>>> bitcoin-dev mailing list
>> 
>>> bitcoin-dev@lists•linuxfoundation.org
>> 
>>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev [1]
>> 
>> _______________________________________________
>> 
>> bitcoin-dev mailing list
>> 
>> bitcoin-dev@lists•linuxfoundation.org
>> 
>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev [1]
>> 
>> _______________________________________________
>> bitcoin-dev mailing list
>> bitcoin-dev@lists•linuxfoundation.org
>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev [1]
> 
> 
> 
> Links:
> ------
> [1] https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
> [2] http://breadwallet.com