One Time Signatures as an Advantage?
7 messages, 20 May to 8 Jun 2026. Started by Jason Resch.
Jason Resch (2), Peter Todd (2), Mikhail Kudinov, Murch, Anthony Towns
NIST is standardizing SLH-DSA as a stateless, post-quantum-secure hash-based signature scheme. However, to achieve the stateless feature of being able to sign multiple messages, requires a significant size overhead.
SLH-DSA (for parameters n=16, w=16) results in signatures that are 7,888 bytes long.
However, if statelessness isn't required, and this can be reduced to 900 bytes for something like XMSS using the same parameters.
Furthermore, if multiple signings per key are dropped as a requirement, and "one time signatures" are used (e.g. WOTS+) then this size reduces further to 560 bytes.
This is a ~14× reduction in signature size for a feature that Bitcoin transactions not only don't need, but are strongly discouraged if not harmful. Using the same key more than once is only required if one is reusing the same address (discouraged), or if one is attempting some kind of double-spend attack.
This could be seen as a sort of advantage: if one attempts to double-spend, they may expose their private key. This same property was an element of Chaum's digital cash: attempting to double-spend exposed you.
Is there any advocacy for NIST to standardize stateful or one-time-use signature algorithms? They seem well-suited to the block-chain use case, where there is always global and persistent state, and keys ought not be re-used. Though this needs to be carefully managed by wallet software: to only expose a one-time-use address to handle a single transaction with a single payer, and never use a OTS address for any kind of public-facing or long-term donation address. Perhaps this complication makes OTS not worth introducing generally, but their space saving properties are attractive.
Jason
Stateful hash-based schemes have been recommended by NIST. See https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-208.pdf . Also you might be interested to read about SHRINCS: https://delvingbitcoin.org/t/shrincs-324-byte-stateful-post-quantum-signatures-with-static-backups/2158 .
Best, Mikhail
On Wed, May 20, 2026 at 7:44 PM Jason Resch <jasonresch@gmail•com> wrote:
Quoted text (45 lines)
> NIST is standardizing SLH-DSA as a stateless, post-quantum-secure > hash-based signature scheme. However, to achieve the stateless feature of > being able to sign multiple messages, requires a significant size overhead. > > SLH-DSA (for parameters n=16, w=16) results in signatures that are 7,888 > bytes long. > > However, if statelessness isn't required, and this can be reduced to 900 > bytes for something like XMSS using the same parameters. > > Furthermore, if multiple signings per key are dropped as a requirement, > and "one time signatures" are used (e.g. WOTS+) then this size reduces > further to 560 bytes. > > This is a ~14× reduction in signature size for a feature that Bitcoin > transactions not only don't need, but are strongly discouraged if not > harmful. Using the same key more than once is only required if one is > reusing the same address (discouraged), or if one is attempting some kind > of double-spend attack. > > This could be seen as a sort of advantage: if one attempts to > double-spend, they may expose their private key. This same property was an > element of Chaum's digital cash: attempting to double-spend exposed you. > > Is there any advocacy for NIST to standardize stateful or one-time-use > signature algorithms? They seem well-suited to the block-chain use case, > where there is always global and persistent state, and keys ought not be > re-used. Though this needs to be carefully managed by wallet software: to > only expose a one-time-use address to handle a single transaction with a > single payer, and never use a OTS address for any kind of public-facing or > long-term donation address. Perhaps this complication makes OTS not worth > introducing generally, but their space saving properties are attractive. > > Jason > > -- > 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/d3648bd4-03d3-4b98-92bf-d845302be349n%40googlegroups.com > <https://groups.google.com/d/msgid/bitcoindev/d3648bd4-03d3-4b98-92bf-d845302be349n%40googlegroups.com?utm_medium=email&utm_source=footer> > . >
On Thu, May 21, 2026, 5:54 AM Mikhail Kudinov <mkudinov@blockstream•com> wrote:
Quoted text (4 lines)
> Stateful hash-based schemes have been recommended by NIST. See > https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-208.pdf > . >
Is the intention that these recommendations will eventually lead to full standardization and FIPS compliance?
Will Bitcoin wait for that do you think, or will it lead in adopting something like SHRINCS first and then NIST will follow in certifying what would by then have become a de facto standard?
Also you might be interested to read about SHRINCS:
Quoted text (3 lines)
> https://delvingbitcoin.org/t/shrincs-324-byte-stateful-post-quantum-signatures-with-static-backups/2158 > . >
I learned of SHRINCS just after making my post. I think it provides the best of both worlds: short signatures in the general case when state is available + the flexibility to fall back and sign many times in case state is lost or the stateful key is used too many times.
In terms of PQC algorithms I'm aware of, this one seems to be among the best in terms of its conservative security assumptions and compactness, and the flexibility to sign many times when necessary addresses the problem with public donation addresses.
Jason
Quoted text (49 lines)
> > On Wed, May 20, 2026 at 7:44 PM Jason Resch <jasonresch@gmail•com> wrote: > >> NIST is standardizing SLH-DSA as a stateless, post-quantum-secure >> hash-based signature scheme. However, to achieve the stateless feature of >> being able to sign multiple messages, requires a significant size overhead. >> >> SLH-DSA (for parameters n=16, w=16) results in signatures that are 7,888 >> bytes long. >> >> However, if statelessness isn't required, and this can be reduced to 900 >> bytes for something like XMSS using the same parameters. >> >> Furthermore, if multiple signings per key are dropped as a requirement, >> and "one time signatures" are used (e.g. WOTS+) then this size reduces >> further to 560 bytes. >> >> This is a ~14× reduction in signature size for a feature that Bitcoin >> transactions not only don't need, but are strongly discouraged if not >> harmful. Using the same key more than once is only required if one is >> reusing the same address (discouraged), or if one is attempting some kind >> of double-spend attack. >> >> This could be seen as a sort of advantage: if one attempts to >> double-spend, they may expose their private key. This same property was an >> element of Chaum's digital cash: attempting to double-spend exposed you. >> >> Is there any advocacy for NIST to standardize stateful or one-time-use >> signature algorithms? They seem well-suited to the block-chain use case, >> where there is always global and persistent state, and keys ought not be >> re-used. Though this needs to be carefully managed by wallet software: to >> only expose a one-time-use address to handle a single transaction with a >> single payer, and never use a OTS address for any kind of public-facing or >> long-term donation address. Perhaps this complication makes OTS not worth >> introducing generally, but their space saving properties are attractive. >> >> Jason >> >> -- >> 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/d3648bd4-03d3-4b98-92bf-d845302be349n%40googlegroups.com >> <https://groups.google.com/d/msgid/bitcoindev/d3648bd4-03d3-4b98-92bf-d845302be349n%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> >
One-time signature schemes are not well-suited for Bitcoin because they:
- cannot be used to participate in multi-user transaction (as another participant could fail the process and force a second signature) - incur lost funds or lost keys upon address reuse (as every node would need to track every output script to prevent duplicates, and the recipient has no say in their output script being sent to another time) - are incompatible with transaction replacement (zero-conf enthusiasts rejoice!)
Murch
On 2026-05-20 10:41, Jason Resch wrote:
Quoted text (45 lines)
> NIST is standardizing SLH-DSA as a stateless, post-quantum-secure > hash-based signature scheme. However, to achieve the stateless feature > of being able to sign multiple messages, requires a significant size > overhead. > > SLH-DSA (for parameters n=16, w=16) results in signatures that are > 7,888 bytes long. > > However, if statelessness isn't required, and this can be reduced to > 900 bytes for something like XMSS using the same parameters. > > Furthermore, if multiple signings per key are dropped as a > requirement, and "one time signatures" are used (e.g. WOTS+) then this > size reduces further to 560 bytes. > > This is a ~14× reduction in signature size for a feature that Bitcoin > transactions not only don't need, but are strongly discouraged if not > harmful. Using the same key more than once is only required if one is > reusing the same address (discouraged), or if one is attempting some > kind of double-spend attack. > > This could be seen as a sort of advantage: if one attempts to > double-spend, they may expose their private key. This same property > was an element of Chaum's digital cash: attempting to double-spend > exposed you. > > Is there any advocacy for NIST to standardize stateful or one-time-use > signature algorithms? They seem well-suited to the block-chain use > case, where there is always global and persistent state, and keys > ought not be re-used. Though this needs to be carefully managed by > wallet software: to only expose a one-time-use address to handle a > single transaction with a single payer, and never use a OTS address > for any kind of public-facing or long-term donation address. Perhaps > this complication makes OTS not worth introducing generally, but their > space saving properties are attractive. > > Jason > -- > 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/d3648bd4-03d3-4b98-92bf-d845302be349n%40googlegroups.com > <https://groups.google.com/d/msgid/bitcoindev/d3648bd4-03d3-4b98-92bf-d845302be349n%40googlegroups.com?utm_medium=email&utm_source=footer>.
On Thu, May 28, 2026 at 10:25:51AM -0700, Murch wrote:
Quoted text (9 lines)
> One-time signature schemes are not well-suited for Bitcoin because they: > > - cannot be used to participate in multi-user transaction (as another > participant could fail the process and force a second signature) > - incur lost funds or lost keys upon address reuse (as every node would need > to track every output script to prevent duplicates, and the recipient has no > say in their output script being sent to another time) > - are incompatible with transaction replacement (zero-conf enthusiasts > rejoice!)
Note that you can design a message signing scheme where you can use a pubkey to sign a merkle tree of messages. In the case of a transaction, multiple conflicting versions of the transaction with different fee rates.
Basically, a One Time Signing Scheme that produces multiple signatures for one signing event.
On Thu, Jun 04, 2026 at 11:01:25AM +0000, Peter Todd wrote:
Quoted text (12 lines)
> On Thu, May 28, 2026 at 10:25:51AM -0700, Murch wrote: > > One-time signature schemes are not well-suited for Bitcoin because they: > > - cannot be used to participate in multi-user transaction (as another > > participant could fail the process and force a second signature) > > - incur lost funds or lost keys upon address reuse (as every node would need > > to track every output script to prevent duplicates, and the recipient has no > > say in their output script being sent to another time) > > - are incompatible with transaction replacement (zero-conf enthusiasts > > rejoice!) > Note that you can design a message signing scheme where you can use a pubkey to > sign a merkle tree of messages. In the case of a transaction, multiple > conflicting versions of the transaction with different fee rates.
There's a balance to be had between what's considered part of the message signing scheme versus what's part of the scripting language.
The scheme above could also be thought of as signing a single message "spend this utxo is valid if this script's conditions are meant", where the script is "or(fee <= 0.5, and(nSequence >= 100, fee <= 1.0), and(nSequence >= 1000, fee <= 30.0))"
Delegating to a signed script like this is essentially the "graftroot" concept.
Cheers, aj
On Fri, Jun 05, 2026 at 05:54:50PM +1000, Anthony Towns wrote:
Quoted text (12 lines)
> On Thu, Jun 04, 2026 at 11:01:25AM +0000, Peter Todd wrote: > > Note that you can design a message signing scheme where you can use a pubkey to > > sign a merkle tree of messages. In the case of a transaction, multiple > > conflicting versions of the transaction with different fee rates. > > There's a balance to be had between what's considered part of the message > signing scheme versus what's part of the scripting language. > > The scheme above could also be thought of as signing a single message > "spend this utxo is valid if this script's conditions are meant", > where the script is "or(fee <= 0.5, and(nSequence >= 100, fee <= 1.0), > and(nSequence >= 1000, fee <= 30.0))"
No, that's a very different scheme. Also, I believe it is vulnerable to miners simply picking the best possible fee rate for them; nSequence isn't big enough to avoid being guessed.
Indeed, I neglected to point out something important in the above: you probably want to include secret nonces (or similar) in each individual branch of the signature tree, to ensure that different versions of it can't be brute forced. Thus to reveal a new option for miners, you make public a previously secret nonce. 128-bits would be sufficiently large to prevent brute-forcing.