SPV clients and relaying double spends
10 messages, 26 to 28 Sep 2014. Started by Aaron Voisine.
Aaron Voisine (3), Matt Whitlock (2), Christophe Biocca, s7r, Tom Harding, Peter Todd, Luke Dashjr
There was some discussion of having nodes relay double-spends in order to alert the network about double spend attempts.
A lot more users will be using SPV wallets in the future, and one of the techniques SPV clients use to judge how likely a transaction is to be confirmed is if it propagates across the network. I wonder if and when double-spend relaying is introduced, if nodes should also send BIP61 reject messages or something along those lines to indicate which transactions those nodes believe to be invalid, but are relaying anyway.
This would be subject to sybil attacks, as is monitoring propagation, however it does still increase the cost of performing a 0 confirmation double spend attack on an SPV client above just relaying double-spends without indicating if a node believes the transaction to be valid.
Aaron Voisine breadwallet.com
What's to stop an attacker from broadcasting millions of spends of the same output(s) and overwhelming nodes with slower connections? Might it be a better strategy not to relay the actual transactions (after the first) but rather only propagate (once) some kind of double-spend alert?
On Thursday, 25 September 2014, at 7:02 pm, Aaron Voisine wrote:
Quoted text (18 lines)
> There was some discussion of having nodes relay double-spends in order > to alert the network about double spend attempts. > > A lot more users will be using SPV wallets in the future, and one of > the techniques SPV clients use to judge how likely a transaction is to > be confirmed is if it propagates across the network. I wonder if and > when double-spend relaying is introduced, if nodes should also send > BIP61 reject messages or something along those lines to indicate which > transactions those nodes believe to be invalid, but are relaying > anyway. > > This would be subject to sybil attacks, as is monitoring propagation, > however it does still increase the cost of performing a 0 confirmation > double spend attack on an SPV client above just relaying double-spends > without indicating if a node believes the transaction to be valid. > > Aaron Voisine > breadwallet.com
Something like that would be a great help for SPV clients that can't detect double spends on their own. (still limited of course to sybil attack concerns)
Aaron Voisine breadwallet.com
On Thu, Sep 25, 2014 at 7:07 PM, Matt Whitlock <bip@mattwhitlock•name> wrote:
Quoted text (23 lines)
> What's to stop an attacker from broadcasting millions of spends of the same output(s) and overwhelming nodes with slower connections? Might it be a better strategy not to relay the actual transactions (after the first) but rather only propagate (once) some kind of double-spend alert? > > > On Thursday, 25 September 2014, at 7:02 pm, Aaron Voisine wrote: >> There was some discussion of having nodes relay double-spends in order >> to alert the network about double spend attempts. >> >> A lot more users will be using SPV wallets in the future, and one of >> the techniques SPV clients use to judge how likely a transaction is to >> be confirmed is if it propagates across the network. I wonder if and >> when double-spend relaying is introduced, if nodes should also send >> BIP61 reject messages or something along those lines to indicate which >> transactions those nodes believe to be invalid, but are relaying >> anyway. >> >> This would be subject to sybil attacks, as is monitoring propagation, >> however it does still increase the cost of performing a 0 confirmation >> double spend attack on an SPV client above just relaying double-spends >> without indicating if a node believes the transaction to be valid. >> >> Aaron Voisine >> breadwallet.com >
Probably the first double-spend attempt (i.e., the second transaction to spend the same output(s) as another tx already in the mempool) would still need to be relayed. A simple "double-spend alert" wouldn't work because it could be forged. But after there have been two attempts to spend the same output, no further transactions spending that same output should be relayed, in order to prevent flooding the network.
On Thursday, 25 September 2014, at 7:12 pm, Aaron Voisine wrote:
Quoted text (32 lines)
> Something like that would be a great help for SPV clients that can't > detect double spends on their own. (still limited of course to sybil > attack concerns) > > Aaron Voisine > breadwallet.com > > > On Thu, Sep 25, 2014 at 7:07 PM, Matt Whitlock <bip@mattwhitlock•name> wrote: > > What's to stop an attacker from broadcasting millions of spends of the same output(s) and overwhelming nodes with slower connections? Might it be a better strategy not to relay the actual transactions (after the first) but rather only propagate (once) some kind of double-spend alert? > > > > > > On Thursday, 25 September 2014, at 7:02 pm, Aaron Voisine wrote: > >> There was some discussion of having nodes relay double-spends in order > >> to alert the network about double spend attempts. > >> > >> A lot more users will be using SPV wallets in the future, and one of > >> the techniques SPV clients use to judge how likely a transaction is to > >> be confirmed is if it propagates across the network. I wonder if and > >> when double-spend relaying is introduced, if nodes should also send > >> BIP61 reject messages or something along those lines to indicate which > >> transactions those nodes believe to be invalid, but are relaying > >> anyway. > >> > >> This would be subject to sybil attacks, as is monitoring propagation, > >> however it does still increase the cost of performing a 0 confirmation > >> double spend attack on an SPV client above just relaying double-spends > >> without indicating if a node believes the transaction to be valid. > >> > >> Aaron Voisine > >> breadwallet.com > >
Of course you wouldn't want nodes to propagate alerts without independently verifying them, otherwise anyone could just issue alerts for every new transaction.
Aaron Voisine breadwallet.com
On Thu, Sep 25, 2014 at 7:16 PM, Matt Whitlock <bip@mattwhitlock•name> wrote:
Quoted text (36 lines)
> Probably the first double-spend attempt (i.e., the second transaction to spend the same output(s) as another tx already in the mempool) would still need to be relayed. A simple "double-spend alert" wouldn't work because it could be forged. But after there have been two attempts to spend the same output, no further transactions spending that same output should be relayed, in order to prevent flooding the network. > > > On Thursday, 25 September 2014, at 7:12 pm, Aaron Voisine wrote: >> Something like that would be a great help for SPV clients that can't >> detect double spends on their own. (still limited of course to sybil >> attack concerns) >> >> Aaron Voisine >> breadwallet.com >> >> >> On Thu, Sep 25, 2014 at 7:07 PM, Matt Whitlock <bip@mattwhitlock•name> wrote: >> > What's to stop an attacker from broadcasting millions of spends of the same output(s) and overwhelming nodes with slower connections? Might it be a better strategy not to relay the actual transactions (after the first) but rather only propagate (once) some kind of double-spend alert? >> > >> > >> > On Thursday, 25 September 2014, at 7:02 pm, Aaron Voisine wrote: >> >> There was some discussion of having nodes relay double-spends in order >> >> to alert the network about double spend attempts. >> >> >> >> A lot more users will be using SPV wallets in the future, and one of >> >> the techniques SPV clients use to judge how likely a transaction is to >> >> be confirmed is if it propagates across the network. I wonder if and >> >> when double-spend relaying is introduced, if nodes should also send >> >> BIP61 reject messages or something along those lines to indicate which >> >> transactions those nodes believe to be invalid, but are relaying >> >> anyway. >> >> >> >> This would be subject to sybil attacks, as is monitoring propagation, >> >> however it does still increase the cost of performing a 0 confirmation >> >> double spend attack on an SPV client above just relaying double-spends >> >> without indicating if a node believes the transaction to be valid. >> >> >> >> Aaron Voisine >> >> breadwallet.com >> >
A lot of this discussion has already occured. Some code was even merged into master, then reverted.
See: https://github.com/bitcoin/bitcoin/issues/4550 https://github.com/bitcoin/bitcoin/pull/4570
It would probably be a good idea to start from that code, as it addresses many of the possible pitfalls you've been discussing.
On Thu, Sep 25, 2014 at 10:37 PM, Aaron Voisine <voisine@gmail•com> wrote:
Quoted text (56 lines)
> Of course you wouldn't want nodes to propagate alerts without > independently verifying them, otherwise anyone could just issue alerts > for every new transaction. > > Aaron Voisine > breadwallet.com > > > On Thu, Sep 25, 2014 at 7:16 PM, Matt Whitlock <bip@mattwhitlock•name> wrote: >> Probably the first double-spend attempt (i.e., the second transaction to spend the same output(s) as another tx already in the mempool) would still need to be relayed. A simple "double-spend alert" wouldn't work because it could be forged. But after there have been two attempts to spend the same output, no further transactions spending that same output should be relayed, in order to prevent flooding the network. >> >> >> On Thursday, 25 September 2014, at 7:12 pm, Aaron Voisine wrote: >>> Something like that would be a great help for SPV clients that can't >>> detect double spends on their own. (still limited of course to sybil >>> attack concerns) >>> >>> Aaron Voisine >>> breadwallet.com >>> >>> >>> On Thu, Sep 25, 2014 at 7:07 PM, Matt Whitlock <bip@mattwhitlock•name> wrote: >>> > What's to stop an attacker from broadcasting millions of spends of the same output(s) and overwhelming nodes with slower connections? Might it be a better strategy not to relay the actual transactions (after the first) but rather only propagate (once) some kind of double-spend alert? >>> > >>> > >>> > On Thursday, 25 September 2014, at 7:02 pm, Aaron Voisine wrote: >>> >> There was some discussion of having nodes relay double-spends in order >>> >> to alert the network about double spend attempts. >>> >> >>> >> A lot more users will be using SPV wallets in the future, and one of >>> >> the techniques SPV clients use to judge how likely a transaction is to >>> >> be confirmed is if it propagates across the network. I wonder if and >>> >> when double-spend relaying is introduced, if nodes should also send >>> >> BIP61 reject messages or something along those lines to indicate which >>> >> transactions those nodes believe to be invalid, but are relaying >>> >> anyway. >>> >> >>> >> This would be subject to sybil attacks, as is monitoring propagation, >>> >> however it does still increase the cost of performing a 0 confirmation >>> >> double spend attack on an SPV client above just relaying double-spends >>> >> without indicating if a node believes the transaction to be valid. >>> >> >>> >> Aaron Voisine >>> >> breadwallet.com >>> > > > ------------------------------------------------------------------------------ > Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer > Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports > Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper > Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer > http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk > _______________________________________________ > Bitcoin-development mailing list > Bitcoin-development@lists•sourceforge•net > https://lists.sourceforge.net/lists/listinfo/bitcoin-development
On 9/26/2014 5:16 AM, Matt Whitlock wrote:
Quoted text (8 lines)
> Probably the first double-spend attempt (i.e., the second > transaction to spend the same output(s) as another tx already in > the mempool) would still need to be relayed. A simple > "double-spend alert" wouldn't work because it could be forged. But > after there have been two attempts to spend the same output, no > further transactions spending that same output should be relayed, > in order to prevent flooding the network. >
This sounds rational - is this already implemented nowadays or *SHOULD BE* implemented to prevent this attack type in the future?
Quoted text (46 lines)
> > On Thursday, 25 September 2014, at 7:12 pm, Aaron Voisine wrote: >> Something like that would be a great help for SPV clients that >> can't detect double spends on their own. (still limited of >> course to sybil attack concerns) >> >> Aaron Voisine breadwallet.com >> >> >> On Thu, Sep 25, 2014 at 7:07 PM, Matt Whitlock >> <bip@mattwhitlock•name> wrote: >>> What's to stop an attacker from broadcasting millions of >>> spends of the same output(s) and overwhelming nodes with >>> slower connections? Might it be a better strategy not to relay >>> the actual transactions (after the first) but rather only >>> propagate (once) some kind of double-spend alert? >>> >>> >>> On Thursday, 25 September 2014, at 7:02 pm, Aaron Voisine >>> wrote: >>>> There was some discussion of having nodes relay >>>> double-spends in order to alert the network about double >>>> spend attempts. >>>> >>>> A lot more users will be using SPV wallets in the future, >>>> and one of the techniques SPV clients use to judge how likely >>>> a transaction is to be confirmed is if it propagates across >>>> the network. I wonder if and when double-spend relaying is >>>> introduced, if nodes should also send BIP61 reject messages >>>> or something along those lines to indicate which >>>> transactions those nodes believe to be invalid, but are >>>> relaying anyway. >>>> >>>> This would be subject to sybil attacks, as is monitoring >>>> propagation, however it does still increase the cost of >>>> performing a 0 confirmation double spend attack on an SPV >>>> client above just relaying double-spends without indicating >>>> if a node believes the transaction to be valid. >>>> >>>> Aaron Voisine breadwallet.com >>> > > ------------------------------------------------------------------------------ > > >
Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer
Quoted text (8 lines)
> Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS > Reports Are you Audit-Ready for PCI DSS 3.0 Compliance? Download > White paper Comply to PCI DSS 3.0 Requirement 10 and 11.5 with > EventLog Analyzer > http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk > > >
On 9/25/2014 7:37 PM, Aaron Voisine wrote:
Quoted text (2 lines)
> Of course you wouldn't want nodes to propagate alerts without > independently verifying them
How would a node independently verify a double-spend alert, other than by having access to an actual signed double-spend?
#4570 relays the first double-spend AS an alert. Running this branch on mainnet, I have been keeping a live list of relayed double-spend transactions at http://respends.thinlink.com
On Sat, Sep 27, 2014 at 07:55:44PM -0700, Tom Harding wrote:
Quoted text (9 lines)
> On 9/25/2014 7:37 PM, Aaron Voisine wrote: > > Of course you wouldn't want nodes to propagate alerts without > > independently verifying them > How would a node independently verify a double-spend alert, other than > by having access to an actual signed double-spend? > > #4570 relays the first double-spend AS an alert. Running this branch on > mainnet, I have been keeping a live list of relayed double-spend > transactions at http://respends.thinlink.com
Speaking of, I ported my replace-by-fee branch the recent v0.9.3 release: https://github.com/petertodd/bitcoin/tree/replace-by-fee-v0.9.3
I actually ported it a few days ago; that release has been running on a half-dozen or so nodes right now for a few days with no issues.
The v0.9.3 release's scriptSig size limit increase adds a new category of double-spending exploit. I'm not going to get time to add that exploit to my replace-by-fee toolkit(1) for at least another week or so though - pull-reqs accepted.
1) https://github.com/petertodd/replace-by-fee-tools
On Sunday, September 28, 2014 5:15:53 AM Peter Todd wrote:
Quoted text (24 lines)
> On Sat, Sep 27, 2014 at 07:55:44PM -0700, Tom Harding wrote: > > On 9/25/2014 7:37 PM, Aaron Voisine wrote: > > > Of course you wouldn't want nodes to propagate alerts without > > > independently verifying them > > > > How would a node independently verify a double-spend alert, other than > > by having access to an actual signed double-spend? > > > > #4570 relays the first double-spend AS an alert. Running this branch on > > mainnet, I have been keeping a live list of relayed double-spend > > transactions at http://respends.thinlink.com > > Speaking of, I ported my replace-by-fee branch the recent v0.9.3 > release: https://github.com/petertodd/bitcoin/tree/replace-by-fee-v0.9.3 > > I actually ported it a few days ago; that release has been running on a > half-dozen or so nodes right now for a few days with no issues. > > The v0.9.3 release's scriptSig size limit increase adds a new category > of double-spending exploit. I'm not going to get time to add that > exploit to my replace-by-fee toolkit(1) for at least another week or so > though - pull-reqs accepted. > > 1) https://github.com/petertodd/replace-by-fee-tools
Do you have or can you provide a version compatible with CPFP, such that a child paying a higher fee trumps the parent's replacement?
Preferably something that will merge cleanly into 0.9.x-ljr :)
Luke