From: Antoine Riard <antoine.riard@gmail.com>
To: conduition <conduition@proton.me>
Cc: Bitcoin Development Mailing List <bitcoindev@googlegroups.com>,
btc@ariard.me
Subject: Re: [bitcoindev] The game-theory problems of PQ sunsetting modes
Date: Mon, 27 Jul 2026 00:45:57 +0100 [thread overview]
Message-ID: <CALZpt+EjD5h9387diQvwtUY9nV-GFgN-Ukp9tY=uqBofb-w7Tg@mail.gmail.com> (raw)
In-Reply-To: <Mh9z4ISxh4iA80hltDJ3_u7ZWSO5pFuMYI9ae0aGIoSXJDdTOi2KiI4Ckm9BfUAY7YEMvTofep7hyKsA7Qjk6ydqk2GqG9fvOzEym9YQKjc=@proton.me>
[-- Attachment #1: Type: text/plain, Size: 17583 bytes --]
Hello Conduition,
> Solid analysis Antoine. However things play out here, activating a PQ
sunset fork of any kind while in the company of a CRQC is apparently quite
hard to do right without setting the incentives up such that they sabotage
the whole effort.
That's the insight that my post aims to underscore, effectively
that activating a PQ sunset fork of any kind might be very hard
in the presence of one or more company with a CQRC, at the very
least there is a lot of uncertainty due to the incentives.
You're correct that the problem appears as soon as you have a
company with one CRQC, where it can just go really deep in the
history of the chain. As soon as you start to have two CRQC,
there is an advantage to burn more EC coins in fees, to reorg to
your advantage, so we're back with some notion of chain finality.
See more comments on rough ideas to alleviate the issue.
> Very neat observation. For such a rollback to occur, the miners would
have to cooperatively elect to stop mining the more mature ("authentic")
chain, where users have already migrated/forked, and instead start mining
on an old block (the "revisionist" chain). Any resources they spend on this
mining will have no payoff until the cumulative proof-of-work of the
revisionist chain surpasses that of the authentic chain. Until then, honest
validator nodes will simply sit idle.
Roughly in what you're describing yes. If you're a miner, I think
you can play even more sneaky chain games on what you're mentioning
about ressources. Let's say you're gaining "coins" on the "authentic"
chain, and after the 100 blocks maturity rule, you immediately short
them on the market to reinvest your proceedings in the energy cost of
the "revisionist" chain (or do a mining halt, as not mining might give
an advantage to the "revisionist" chain).
You're a miner, if the "authentic" chain wins, that's fine you're
already re-sell the matured coin. If the "revisionist" chain wins,
you got new fresh coinbase on the "revisionist" chain i.e a double-spend,
plus any CQRC "bounty" coming from the exploited coins.
I think there is some "mining silent reorg" advantage here.
> Due to the vast incentive towards colluding with the CRQC, maybe this
would be feasible for some large miners?
>
> This would essentially be a massive double-spend attack as well, since
miners who successfully roll back the blockchain in this way would be
reorging their own mining earnings out of existence, some of which they
presumably sold (on the authentic chain) to pay for electricity. This might
make the exchanges they sold the coins to extremely unhappy: The miners are
effectively retconning their own deposits.
See my point above, they're only limited by the 100 block maturity rules.
Yes, miners would start to be unable to settle "fresh" coins, but also
"old"
coins (both EC and PQ), as they are at risk of being roll back (at least for
the ones who are weeks recent).
> For this to happen, miners must be able to withstand significant capex
(on mining a revisionist chain), while being blackballed by exchanges, and
possibly also devaluing the very coins they were bribed with by the CRQC.
And even then, it's not clear how - assuming they were able to pull the
attack off and remain solvent - the miners would actually use the
ill-gotten coins, and whether they'd have any value on the other side of a
successful deep reorg attack.
One of the point of the analysis, they might be able to fund their capex,
with
the exploited coins, if they can get liquidity for them on the exchanges.
On the
other hand, as long as they're economically solvent, they can keep the
EC-exploited
coins, until some far future, when the market price them at an interesting
price
enough, and slowly and covertly sell them out of their balance sheet by
then.
> Still, for shallow reorgs (a few blocks) this seems like a worthwhile
concern that seriously hampers any tripwire attempts. The best case is if
we can deploy the EC disabling fork before such tempting incentives enter
the field of play.
A naive tripwire (there might be more secure design worthy to think more
about it),
of course sounds it will be always at risk of being keep out of the chain
by miners.
Even the "we move fast and try to disable EC", assuming it's
philosophically acceptable
by the community, and I'm among the one disagreeing to do so, as I pointed
out in my
previous post, the stack of "lost" EC coins might be worth 10 years of
bitcoins, that's
a lot of reorg budget.
So in the hypothesis, you have a "sunset" activation, then 3 months after a
CRQC
getting out of the box, there is a lot of incentives uncertainty. Even
worst, there
is even a "Lorentz effect", where the public and well-known "sunset"
activation,
incentives "advanced adversaries" to reveal the CRQC they were keeping
sleeping
in the backyard, as they know after the activation the cost structure is
altered.
I'm thinking there are other solutions that we have not explored yet such as
PQ-blessed periodic checkpoints. E.g, let's say that every month, by
consensus
rule, there is a checkpoint published that needs to be finalized e.g being
signed by more than % of PQ-safe pubkeys (e.g %1 of the overall coins).
As those pubkeys are PQ-safe, they cannot be forged by a CRQC, and a mining
coalition would not be able to go deeper than this checkpoint height,
capping
up the maximum of EC-unsafe coins that could be used as reorg budget. Of
course,
that would ask for a number of stakeholders in the ecosystem to have online
keys
for the "checkpoint" finalization, though that % can be kept low [0] [1].
It's just a "rough idea", as somehow it's re-introducing a form of
checkpoint
(which is meehhhh...but not worse than the sun setting ideas imho). I think
there
are more imaginative ideas that we can come up on the design table to
alleviate
the risk of a CQRC acting in coordination with a mining coalition.
Notwithstanding the ultimate direction taken, the first primitive that
would need it to give us more design flexibility would be the PQ-safe
signing
algorithm, be it Falcon, SkiSign or whatever.
Best,
Antoine
OTS hash: bdb435e6c81eb762bb6a12a03e2c23a83396a3481b8894af79732c2e2b569682
[0] This also let the door open for a EC coins owner, of which the pubkey
has not been revealed e.g P2TR to exfiltrate their old coins towards
safer one.
[1] The checkpoint would have to be carefully designed to avoid being
tampered
by a miner, e.g a PQ-signed checkpoint would gain a proof-of-work bonus
discount ?
I'm already far in the territory of heretical consensus design ideas...
Le dim. 19 juil. 2026 à 23:02, conduition <conduition@proton.me> a écrit :
> Solid analysis Antoine. However things play out here, activating a PQ
> sunset fork of any kind *while in the company of a CRQC* is apparently
> quite hard to do right without setting the incentives up such that they
> sabotage the whole effort.
>
> If a CQRC entity is able to build a coalition with a 51% majority of
> miners, the "upgraded
> PQ safe" coins might be also at risk [4]. Indeed, such malicious coalition
> could just roll-back the chain state back to the migration height of
> said coin, solve the DL for this coin and unroll back forward the chain.
>
> I do not believe that the old chain history would be safe from deep
> reorgs attacks by CQRC capable entities, as soft-fork deployments are
> "height-based" burnt and not "hash-based" burnt (BIP90). Checkpoints
> have been removed from the latest bitcoind versions. Maybe user-activated
> checkpoints or other similar mechanisms might be a more robust defense
> against CQRC entities attacking the chain finality.
>
>
> Very neat observation. For such a rollback to occur, the miners would have
> to cooperatively elect to stop mining the more mature ("authentic") chain,
> where users have already migrated/forked, and instead start mining on an
> old block (the "revisionist" chain). Any resources they spend on this
> mining will have no payoff until the cumulative proof-of-work of the
> revisionist chain surpasses that of the authentic chain. Until then, honest
> validator nodes will simply sit idle.
>
> Due to the vast incentive towards colluding with the CRQC, maybe this
> would be feasible for some large miners?
>
> This would essentially be a massive double-spend attack as well, since
> miners who successfully roll back the blockchain in this way would be
> reorging their own mining earnings out of existence, some of which they
> presumably sold (on the authentic chain) to pay for electricity. This might
> make the exchanges they sold the coins to extremely unhappy: The miners are
> effectively retconning their own deposits.
>
> For this to happen, miners must be able to withstand significant capex (on
> mining a revisionist chain), while being blackballed by exchanges, and
> possibly also devaluing the very coins they were bribed with by the CRQC.
> And even then, it's not clear how - assuming they were able to pull the
> attack off and remain solvent - the miners would actually *use* the
> ill-gotten coins, and whether they'd have any value on the other side of a
> successful deep reorg attack.
>
> Still, for shallow reorgs (a few blocks) this seems like a worthwhile
> concern that seriously hampers any tripwire attempts. The best case is if
> we can deploy the EC disabling fork *before* such tempting incentives
> enter the field of play.
>
> regards,
> conduition
> On Sunday, July 12th, 2026 at 12:13 PM, Antoine Riard <
> antoine.riard@gmail.com> wrote:
>
> Hi list,
>
> In this post, I'm extending on the game-theory problems
> underscored for my answer to [ ] to other post-quantum
> sunsetting scenarios previously mentioned on this list.
>
> Firstly, let's remember the "tripwire" idea [0]. With the
> "tripwire", if I understand it correctly we introduce a
> consensus level proof of quantum computers e.g with a NUMS
> puzzle.
>
> This NUMS is committed in a honeypot UTXO let's say with
> some non-null bitcoin reward to unlock it. When the NUMS
> point is solved by a QC entity, it automatically triggers
> a "freeze" of all the "legacy" coins starting at some block
> height-defined window in the future.
>
> While it appears feasible engineering-wise, the problem
> is more on the game-theory plane of analysis. As it was
> previously noted by another commentator than me [1], why
> an economically-rational CRQC entity would go to trigger
> such an evident "honeypot" UTXO depriving it from further
> (covert) extractions of the legacy coins to a safe wallet
> owned by this entity.
>
> A more sophisticated scenario, that I was laying out more
> recently, a 51% majority coalition of miners could coordinate
> with a CRQC entity to censor the transaction inclusion of any
> PQ proof, even an inclusion attempt of a PQ proof generated by
> an honest PQ entity [2].
>
> Exposing again the economic analysis, a year of mining income
> is evaluated at around $20B. The number of legacy P2Pk coins
> is evaluated to be around 1.7 M of coins or as of today $107B.
> If we go to account the numbers of "coin loss", the estimated
> number can be more around 3-4 M, so let's say $215B worth of
> target coins (a coin lost to you is not a coin lost to a CRQC
> entity...).
>
> That's something like ~10 years of potential income, that an
> economically rational miner might not refuse if a miner has
> a credible odd of capturing a share of this magic income to
> the prorata of their hashrate capabilities [3].
>
> If we assume a PQ coin extraction game with 2 CQRC entities
> availing roughly the same capabilities, they might compete for
> the majority hashrate of the miners, those miners solely driven
> by economic incentives. The focal point of equilibrium between
> the two strategies is likely going to be the marginal energy cost
> to run a CQRC, assuming that in a fee race a CQRC entity can
> offer to the majority of miners to burn more of a coin value
> as reorg fee.
>
> Secondly, for the second approach of sunsetting, the one very roughly
> described in BIP361 and based on pure "flag-day" activation, the
> security analysis can extend to this approach too. Even assuming
> a week-long period for a BIP9-like activation mechanism, a coalition
> of miners might stil go to reorg in depth the chain before the
> activation of said soft-fork.
>
> Such an approach is only theoretically increasing the coordination cost
> (and one would observe the asymmetry of information is selecting a time
> horizon period, as a CRQC might appear at any time during this period).
>
> One can observe that the 2 sunsetting approach, be it "tripwire" or
> "flag-day" approaches are introducing a "choke point" to the chain
> finality,
> as in the lack of it a CQRC entity might covertly exfiltrate "legacy"
> coins,
> with no knowledge of the miners, or even without coordination with them.
>
> After the "choke point", a CQRC entity might alter its strategy of going
> overt and start to offer fee bounties to reorg the chain as it's advantage
> to the majority of miners (to not loss an exploitation advantage to another
> CQRC entity).
>
> Finally, in this analysis we're only underscoring the risk of "legacy"
> coins, i.e coins that would have not upgraded to a PQ safe format, after
> some time horizon. However, in the Bitcoin blockchain world, time is
> relative, or rather only thermodynamically convergent. If a CQRC entity
> is able to build a coalition with a 51% majority of miners, the "upgraded
> PQ safe" coins might be also at risk [4]. Indeed, such malicious coalition
> could just roll-back the chain state back to the migration height of
> said coin, solve the DL for this coin and unroll back forward the chain.
>
> I do not believe that the old chain history would be safe from deep
> reorgs attacks by CQRC capable entities, as soft-fork deployments are
> "height-based" burnt and not "hash-based" burnt (BIP90). Checkpoints
> have been removed from the latest bitcoind versions. Maybe user-activated
> checkpoints or other similar mechanisms might be a more robust defense
> against CQRC entities attacking the chain finality.
>
> Current bitcoin mining process and the chain finality is assumed to be
> reasonably secure under the Gambler's Ruin Problem and some other
> assumptions
> (e.g a reliable network to relay the blocks). It might be considered that
> the introduction of CQRC computers might not be only a risk for the
> "legacy"
> coins, though far more concerning for the chain finality itself.
>
> Independently of being philosophically "pro" or "contra" in freezing
> legacy coins, I do believe the irruption of one or more CQRC entities
> and the potential of disruptions on the Bitcoin network stability is
> a subject deserving a bit more research and more work from the development
> community [5].
>
> Cheers,
> Antoine
> OTS hash: 496d9c26c6f3d805dae88f487600f46990572fc84c4ca907fb85b5441c235cf3
>
> [0] https://groups.google.com/g/bitcoindev/c/8O857bRSVV8/m/8nr6I5NIAwAJ
> [1] https://groups.google.com/g/bitcoindev/c/8O857bRSVV8/m/7uu4dZNgAwAJ
> [2] One might consider the following realistic scenario, it
> might that even if a CRQC become relevant, at first it will
> be only operated by big companies let's say in the US or China
> and they will prefer to keep the existence of such capabilities
> hidden for a while for non-economical reasons.
> Suddenly, one of the actor starts to use those post-quantum
> capabilities and the social equilibrium does not hold anymore
> with impactful second-order implications for the Bitcoin ecosystem.
> [3] On the low time incentive miner hypothesis, one can empirically
> observe (as of June '26) than it has limits given how fast are ready
> mainstream mining companies to reallocate their data centers and
> sources of energies to more generic high-performance computations
> rather than SHA256 hashing.
> [4] For the degree of scientificity of "game-theory" in itself, I can
> only forward the reader to the "Formulation of the Economic Problem"
> chapter in the "Theory of Games and Economic Behavior" book from Von
> Neumann & Morgenstern, 1944
> [5] As quantum raises a number of skeptical eyebrows in the community,
> the first elaboration of quantum physics have been as old as the 30's,
> and so far no one has got a Nobel Prize, or any other major scientific
> prize to prove the physical impossibility of a large-scale quantum computer
>
> --
> 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/CALZpt%2BFOUJF3E7YDk5xh-Cv9kxduGiuOPVK5x171%3D25C3ryJPQ%40mail.gmail.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/CALZpt%2BEjD5h9387diQvwtUY9nV-GFgN-Ukp9tY%3DuqBofb-w7Tg%40mail.gmail.com.
[-- Attachment #2: Type: text/html, Size: 21542 bytes --]
prev parent reply other threads:[~2026-07-26 23:53 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-07-12 18:34 [bitcoindev] The game-theory problems of PQ sunsetting modes Antoine Riard
2026-07-19 22:02 ` 'conduition' via Bitcoin Development Mailing List
2026-07-26 23:45 ` Antoine Riard [this message]
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='CALZpt+EjD5h9387diQvwtUY9nV-GFgN-Ukp9tY=uqBofb-w7Tg@mail.gmail.com' \
--to=antoine.riard@gmail.com \
--cc=bitcoindev@googlegroups.com \
--cc=btc@ariard.me \
--cc=conduition@proton.me \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
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