Hi waxwing, > That is true but it is also directly addressed with a mitigation in the section on the attack in BIP78; (already linked here but just to repeat: https://github.com/bitcoin/bips/blob/master/bip-0078.mediawiki#user-content-span_idprobingattackspanOn_the_receiver_side_UTXO_probing_attack ) > specifically it says "When the receiver detects an original transaction being broadcast, or if the receiver detects that the original transaction has been double spent, then they will reuse the UTXO that was exposed for the next payjoin.". > In the blog post on substack, I don't see any discussion of whether the mitigations proposed make sense. I think they do. The [mitigations][0] are not good enough: "While the exposed UTXO will be reused in priority to not leak other UTXOs, there is no strong guarantee about it. This prevents the attacker from detecting with certainty the next payjoin of the merchant to another peer." > One other important thing that is discussed in BIP78, there is a difference between a "merchant" (or in any case, payment-receiving-server) case vs. a peer to peer payments case. In the latter case you cannot simply continuously ask for more and more "invoices" (payjoin urls) from the counterparty. An attacker could use social engineering and ask for multiple payjoin URI or use different nyms. The adversaries I expect to do this are North Korean hackers and government agencies. > I mean, obviously, I don't agree with that. Personally, I would never use payjoin to accept bitcoin payments or donations from untrusted people. > And then there's the 10000ft view: if an attacker doesn't mind spending coins, they can just .. do sender-side actual payjoins, over and over, to try to collect utxos. Agree. > I think that whether this is a problem or not is deeply tied to that inevitable conflict between the desire for privacy and the desire for scalability/low cost. To never co-spend utxos means a wallet fragments to infinite utxos. > But to cospend maximally (as a naive merchant *might* do - receive 1000 payments then send all of them at the same time into a cold wallet) wipes out, at least most of the time, the privacy gain from not reusing addresses in the first place. I don't think every bitcoin merchant could be a victim of this attack. However, if you are selling something illegal and accept bitcoin payments using payjoin, then government agencies might be interested in UTXO probing attacks. > To "read" your current wallet, he somehow has to pay for a huge range of different amounts to try to entice you to spend *different* utxos; it's not easy, but equally it's ridiculous to claim that it doesn't leak anything. Payjoin makes things easier for the attacker. If this attack wasn't a real problem, I don't think it would have been mentioned as a significant drawback in the latest PDK [blog post][1]. [0]: https://github.com/bitcoin/bips/blob/04b2ec649b2201683e63f651d67d30ee4c1f1a28/bip-0078.mediawiki?plain=1#L377 [1]: https://payjoindevkit.org/2025/03/18/the-evolution-of-payjoin/#the-limitations-of-two-party-payjoins /dev/fd0 floppy disk guy On Sat, Mar 29, 2025 at 1:15 AM waxwing/ AdamISZ wrote: > Hi /dev/fd0 and all, > > > The original transaction can be replaced by the attacker, and it would > only cost a few hundred sats or nothing if it's payjoin transaction. I > think such attacks could still be effective if the attacker has the budget > and motivation to spy on someone's wallet. > > That is true but it is also directly addressed with a mitigation in the > section on the attack in BIP78; (already linked here but just to repeat: > https://github.com/bitcoin/bips/blob/master/bip-0078.mediawiki#user-content-span_idprobingattackspanOn_the_receiver_side_UTXO_probing_attack > ) > > specifically it says "When the receiver detects an original transaction > being broadcast, or if the receiver detects that the original transaction > has been double spent, then they will reuse the UTXO that was exposed for > the next payjoin.". > > However it is unclear whether that's part of the protocol specification, > or whether it should be. So there's at least something to discuss there. > > In the blog post on substack, I don't see any discussion of whether the > mitigations proposed make sense. I think they do. Also that blog post > discusses altering the sender code to prevent sending the tx. An > implementation might differ, but at least in BIP78 it should be the > receiver who broadcasts the initial non-payjoin version after a short > timeout. That difference connects with the next point also: > > One other important thing that is discussed in BIP78, there is a > difference between a "merchant" (or in any case, payment-receiving-server) > case vs. a peer to peer payments case. In the latter case you cannot simply > continuously ask for more and more "invoices" (payjoin urls) from the > counterparty. In the former case, you certainly can, and the mitigations > mentioned make sense there to prevent the "utxo collection" algorithm of > continuously failing to complete or double spending, across multiple > payment amounts. > > > It was costless in the demo which could be fixed by bullbitcoin > > Yeah I see a couple of related things there, BIP77 is more nuanced on the > "receiver broadcasts original after short delay" saying that an expiration > MAY be set and applied by receivers, which relates ack to the earlier point > that for a p2p payment arranged with both parties live is not exposed to a > repeated request attack, hence it may not be needed. I do think it comes > back to the "don't change the currently available utxo until it gets used" > statement in that BIP78 section mentioned above. > > With that nuance even your modified-code-sender could be argued not to be > an issue, though I think I prefer the BIP78 inclusion of "receiver > broadcasts after an expiration" being a requirement, not a "MAY". > > > Payjoin should only be used with trusted senders. > > I mean, obviously, I don't agree with that. > > And then there's the 10000ft view: if an attacker doesn't mind spending > coins, they can just .. do sender-side actual payjoins, over and over, to > try to collect utxos. After all the very first blockchain analysis paper by > Meiklejohn et al focused on exactly this; see how much info you can get by > actually paying at a merchant. > > I think that whether this is a problem or not is deeply tied to that > inevitable conflict between the desire for privacy and the desire for > scalability/low cost. To never co-spend utxos means a wallet fragments to > infinite utxos. But to cospend maximally (as a naive merchant *might* do - > receive 1000 payments then send all of them at the same time into a cold > wallet) wipes out, at least most of the time, the privacy gain from not > reusing addresses in the first place. A payjoin based merchant flow, in the > simplest version, would end up linking the 1000 anyway (think a chain of > 1000 payjoins with 1 merchant in and 1 merchant out), but with > substantially more deniability/lack of certainty at each step in terms of > both utxo and amount, and never being hit with "huge transaction during fee > spike". It should at least be *better*. > > If those 1000 payjoins were an attacker, he only really learns about the > first utxo, if you snowball like that. To "read" your current wallet, he > somehow has to pay for a huge range of different amounts to try to entice > you to spend *different* utxos; it's not easy, but equally it's ridiculous > to claim that it doesn't leak anything. > > Cheers, > AdamISZ/waxwing > > On Thursday, March 27, 2025 at 9:19:38 AM UTC-3 /dev /fd0 wrote: > >> Hi jbesraa, >> >> > While the possibility of UTXO probing via Payjoin is a valid concern >> regarding privacy, it's important to note that it might not always come >> without cost for the attacker. The Payjoin recipient > needs to validate >> the initial request, ensuring the sender's inputs are broadcastable. This >> means the recipient could, in practice, broadcast the initial transaction >> even if the sender aborts the > Payjoin. >> >> > Furthermore, implementing strategies like maintaining a set of 'seen >> inputs' can make such probing attempts more easily detectable and less >> effective. >> >> The original transaction can be replaced by the attacker, and it would >> only cost a few hundred sats or nothing if it's payjoin transaction. I >> think such attacks could still be effective if the attacker has the budget >> and motivation to spy on someone's wallet. >> >> /dev/fd0 >> floppy disk guy >> >> >> On Wed, Mar 26, 2025 at 11:54 PM jbesraa wrote: >> >>> While the possibility of UTXO probing via Payjoin is a valid concern >>> regarding privacy, it's important to note that it might not always come >>> without cost for the attacker. The Payjoin recipient needs to validate the >>> initial request, ensuring the sender's inputs are broadcastable. This means >>> the recipient could, in practice, broadcast the initial transaction even if >>> the sender aborts the Payjoin. Furthermore, implementing strategies like >>> maintaining a set of 'seen inputs' can make such probing attempts more >>> easily detectable and less effective. While these measures don't eliminate >>> the privacy considerations entirely, they do highlight that recipients have >>> potential defenses and that probing isn't necessarily a risk-free endeavor >>> for the attacker. >>> >>> On Tuesday, March 25, 2025 at 1:48:15 PM UTC+2 /dev /fd0 wrote: >>> >>> Hi everyone, >>> >>> Sometimes we are curious and want to know about UTXOs in other wallets. >>> Payjoin allows you to do this and the recipient would never doubt it >>> because it's a privacy tool. It's possible to find UTXO in recipient's >>> wallet without sending any bitcoin. It's called UTXO probing attack and >>> described in BIP 77-78. >>> >>> I have shared a demo with all the details in this [post][0]. I have used >>> bullbitcoin wallet for testing this because it was the only [wallet][1] >>> which supports payjoin v2 (send, receive) and testnet3. >>> >>> I think users should be aware of this tradeoff and the information they >>> share with the sender in payjoin. Payjoin should only be used with trusted >>> senders. >>> >>> [0]: >>> https://uncensoredtech.substack.com/p/utxo-probing-attack-using-payjoin >>> [1]: https://en.bitcoin.it/wiki/PayJoin_adoption >>> >>> /dev/fd0 >>> floppy disk guy >>> >>> -- >>> You received this message because you are subscribed to the Google >>> Groups "Bitcoin Development Mailing List" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to bitcoindev+...@googlegroups.com. >>> To view this discussion visit >>> https://groups.google.com/d/msgid/bitcoindev/1c7130d4-cbac-4404-968c-9eb7b4e2e4cbn%40googlegroups.com >>> >>> . >>> >> -- > You received this message because you are subscribed to the Google Groups > "Bitcoin Development Mailing List" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to bitcoindev+unsubscribe@googlegroups.com. > To view this discussion visit > https://groups.google.com/d/msgid/bitcoindev/d0a0e344-d777-49bc-8b3c-a3462f0d6836n%40googlegroups.com > > . > -- You received this message because you are subscribed to the Google Groups "Bitcoin Development Mailing List" group. To unsubscribe from this group and stop receiving emails from it, send an email to bitcoindev+unsubscribe@googlegroups.com. To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/CALiT-Zr78wh7YTE_qi_wotm-84zShDSPth6W4aaaYNMbLGO0sw%40mail.gmail.com.