From mboxrd@z Thu Jan 1 00:00:00 1970 Delivery-date: Mon, 02 Jun 2025 16:03:36 -0700 Received: from mail-qv1-f55.google.com ([209.85.219.55]) by mail.fairlystable.org with esmtps (TLS1.3) tls TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256 (Exim 4.94.2) (envelope-from ) id 1uMEBp-0002Gf-RN for bitcoindev@gnusha.org; Mon, 02 Jun 2025 16:03:35 -0700 Received: by mail-qv1-f55.google.com with SMTP id 6a1803df08f44-6fad5f1e70fsf63903006d6.0 for ; Mon, 02 Jun 2025 16:03:33 -0700 (PDT) ARC-Seal: i=2; a=rsa-sha256; t=1748905407; cv=pass; d=google.com; s=arc-20240605; b=IKhyzVw3Z8zfFXcm///rW3vIx51bXpczZ/4KcrI7QQxQ5EhVVlofNBkXR6ujaApyHQ PJw1+07T4cVgDF+/2FuJ3yNFx4rkzUENoOe1rMUcIDW8+778uwprgAXPzx7o2b6HduVO 351aXnJt1Y56s6bOXLV1BUu0kUhnrV5MlSRkn86uOsMFDKkflWFcZmebyWWjGYL9truY gg5DBYFlw0F4pbxWeCnp1QuHQbHhnhXbM/67WhHYi+ydJ8JzkhjOF3o54j+UiH59S/FH lLM9VoyT8HDkSa3duEOOP6+bqRL562UTlriExmKkFJYsS2NKffbULEWTph8+zPlaccEQ Fikw== ARC-Message-Signature: i=2; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20240605; h=list-unsubscribe:list-subscribe:list-archive:list-help:list-post :list-id:mailing-list:precedence:content-transfer-encoding:cc:to :subject:message-id:date:from:in-reply-to:references:mime-version :sender:dkim-signature:dkim-signature; bh=oJxSF4UR2ztcIG6wInosAQsFW9NCsVAM55HWiSpRXvc=; fh=PPte0pqdA1818NS9U2Lj7lYCTJrnGX0NcCgsdi0lrTk=; b=TzCHaxKdjjsN/643vEWUattcb/nHFV+U//DlD6WbY5sNIisZboRsmuXtETadWzKV5V tXn/ewSbuImsUHnJTnr7QrF851OBEbbr1IKC/P36dC3OFC768aCbJEbdpNljUkVdye4W WB3xRdpl21b1IRgPWQ0Ahz7lzQ/Lo3MoYgMNnhrxUr8wuSbYF/MSw1thXUe6hR4OROkO /emDz1sURU1pHIJ6Gxf2YLIHm94BYKrAk/tRHSwubwHx06Kd5VAGdoemIu3Q9u7oW5B4 kSD5gczcKi2GbNJCSW8vGBx+FU+O0JXLHAx6Ppcz2kYRV3QSIrarDtdsYFqypOGbsGLD eEHg==; darn=gnusha.org ARC-Authentication-Results: i=2; gmr-mx.google.com; dkim=pass header.i=@gmail.com header.s=20230601 header.b=fgMHf7lE; spf=pass (google.com: domain of bnagaev@gmail.com designates 2607:f8b0:4864:20::436 as permitted sender) smtp.mailfrom=bnagaev@gmail.com; dmarc=pass (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com; dara=pass header.i=@googlegroups.com DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=googlegroups.com; s=20230601; t=1748905407; x=1749510207; darn=gnusha.org; h=list-unsubscribe:list-subscribe:list-archive:list-help:list-post :list-id:mailing-list:precedence:x-original-authentication-results :x-original-sender:content-transfer-encoding:cc:to:subject :message-id:date:from:in-reply-to:references:mime-version:sender :from:to:cc:subject:date:message-id:reply-to; bh=oJxSF4UR2ztcIG6wInosAQsFW9NCsVAM55HWiSpRXvc=; b=ti7bQF6sf/DtHJAAX2K86VVmgz6dP36gTDGi/JyfuW93tD6jLR0C2SbvZOsNKnOuhX JI+gjkSHo4FijhGUn260oOreiWBgr22YtslSsBIJjZ0iDrg3FbQSimIa3W3MhmDxV4w2 sjP5z8edOIJeULhPcbNOnFbMQYhoYbxTErQB9BPuKln31ujzI9S450d05jcNfKLIYrUL pfnldysVtdfyuPMD3BIW5q9QC2qmQPiF1gO1hKYL6COIdnWCy1r4DocZxMjcV+k9djjd 2Cfzq41sQND/lsWzW+1O9qXirdX6d7u5v+rjnOHTeg40SWBsIBnGL18J4/tWck9bMAdy QiRA== DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20230601; t=1748905407; x=1749510207; darn=gnusha.org; h=list-unsubscribe:list-subscribe:list-archive:list-help:list-post :list-id:mailing-list:precedence:x-original-authentication-results :x-original-sender:content-transfer-encoding:cc:to:subject :message-id:date:from:in-reply-to:references:mime-version:from:to:cc :subject:date:message-id:reply-to; bh=oJxSF4UR2ztcIG6wInosAQsFW9NCsVAM55HWiSpRXvc=; b=jrbQQO9vkTuYSN4x+rqHWUkFnHbZtoFOFwW0DR3GdyVwjRrsX/9YzS8EaIBBDG8t28 K18NxMgGfkbpk1oh+g8R5Iy1iyEqbdfLWxCcAtam9QwNV/kVEbYUmt+Vysott0BPD5pL 1Zqo1BKNo9zBHkBb2D+S5ORaKlnvM3GWhjQlRohLo5pwlzVxBD4VGzdUiSp5bpdMz+kV YXsq0Iiswyk597YGDDhVPKqImZX2yUtRCCizWwtCtWQTadon6FisNNPyC5c5aeTxAVsi KWDqLTbFU9n6naVaFuC2CbZ6Eq4IrzFDN25DxMyHvMh7BKMYbyF3RvRzonjhrbpAZO9x 1pkw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1748905407; x=1749510207; h=list-unsubscribe:list-subscribe:list-archive:list-help:list-post :list-id:mailing-list:precedence:x-original-authentication-results :x-original-sender:content-transfer-encoding:cc:to:subject :message-id:date:from:in-reply-to:references:mime-version :x-beenthere:x-gm-message-state:sender:from:to:cc:subject:date :message-id:reply-to; bh=oJxSF4UR2ztcIG6wInosAQsFW9NCsVAM55HWiSpRXvc=; b=fkaiHeRPhkhTt+UnYcjNdXtdRQ2y7Sxc886H2iT+mtIVcUly3PZmvHmJCH+NMQUZ2S cv4XdaEdpKNzBgnKe/yvCZbG4pBMJ0QL5v2KD6+fwxSMu5/etRyeMtJ1D9jXtfIrqnEF Iamf7/d4YLbJeCRavet3sH0rYu0zBf0ShKly+MOcVe2xhCP0RrZEsiaevkapUchgnjGj cC/VT9VqPHjOcZ4JkzbVfeK1GvXOCS7qAXoiEzv3ExHUWXnhHv1E2/dqwdNDEMmlg3rL osIltyu4Kj1jqvcadHVhPrN+Ml/YDfYRnVOf9Biisl+gTa3s5ZcI+5LvzFGa3XDHWWJ7 DgAA== Sender: bitcoindev@googlegroups.com X-Forwarded-Encrypted: i=2; AJvYcCVMflX07cQGrznSPLUktakJDTKTd8N0xunBEzL4dvF484pUSfbm+WRxFQOEwcL+SSuxX9Tsr2Gygu2o@gnusha.org X-Gm-Message-State: AOJu0YxPjrTIOC053UCCkaOCjlRESb05hp1hxGxjl9wl3NvBGy6cJgei 5LZnR77P8yIa4DnQUf3yZ01HgH0W53/+urXGnID01WvHjLzitjcUtUn2 X-Google-Smtp-Source: AGHT+IHxS4Gf4MrPtlTUua2Z6iF803+PWbwi+gqZ8iiLaQdB4ZmH8yfg+O4suw6AF31S3y0VgzBmGA== X-Received: by 2002:a05:6214:4008:b0:6fa:ab49:4482 with SMTP id 6a1803df08f44-6fad196721amr210765236d6.35.1748905407239; Mon, 02 Jun 2025 16:03:27 -0700 (PDT) X-BeenThere: bitcoindev@googlegroups.com; h=AZMbMZdknRZz+lvihbCfwP4RIqQITJUk4ydRLTy4Szhyv2D0Cw== Received: by 2002:a05:6214:564c:b0:6fa:bd2a:c680 with SMTP id 6a1803df08f44-6fac5d8a8c9ls97508316d6.2.-pod-prod-06-us; Mon, 02 Jun 2025 16:03:22 -0700 (PDT) X-Received: by 2002:a05:620a:3f42:b0:7d2:fd6:c9fd with SMTP id af79cd13be357-7d20fd6ca2emr200587985a.42.1748905402418; Mon, 02 Jun 2025 16:03:22 -0700 (PDT) Received: by 2002:a05:620a:27cc:b0:7c5:3b15:3956 with SMTP id af79cd13be357-7d210d61645ms85a; Mon, 2 Jun 2025 15:51:03 -0700 (PDT) X-Received: by 2002:a05:620a:2b82:b0:7d0:294a:523c with SMTP id af79cd13be357-7d0a4e3cc77mr2215971285a.41.1748904662210; Mon, 02 Jun 2025 15:51:02 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1748904662; cv=none; d=google.com; s=arc-20240605; b=Q+sQ6gw3oChxY1dCACgr+Kk3fLFX7uqUtEaKzW168WLVGz1x/L7Z7mRAdACCGRpTMA ixc/pWvkLdpsMMy++RkuWmSl0BOctUWhzbfiO2yEQ1sguapYF4WJPviBn+lkN93npii7 HPSYmjldWXPYBJEEM+gNHXIkriorEcAEaj4+z6c94UDdOwqZpp7Ez+qDyHNUVWy7DSfV 3iH3VIwhg/1621zuWDWmYc4sBWZ3NokDt/kzuk0SRYVbhHG70+H+gG+6Q0RKjupgfV9n Lkbr84meOuvbdHwndFS2vMMTDUrS6JfJB6YPc4WyGbDzyx/EhpjogMJSMlO8Wr1fPHPa ExUg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20240605; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:dkim-signature; bh=w2pHXjrWR2R/pV1oMdGsNMmiOynOOxrnycKapixjq/Y=; fh=zfrgopOiIZUyW/QOQMW3tlf+B0T0p2fpIoXRbxgb3Y8=; b=Eu5e91CrtCqJy5IIcm2aKQm3n69zYE9TkfZMSx+D/dEmjBEmE6MPIdy3LYuaFuk39+ x9jX7HylK/h9f3YYHFkm8TNe+N0wqbUSQDLCZ/BowY7zsSnLQhRp4BZJ17+MRtsqioVo vVFyVqyt+xHmmyxCe9d8W7aGoc9JLOMLA+qLx0YZ7sA4REqb2vYVz82ndCrUYoN7ufHR 8JfJAcYgBy1KoiC126ySLrpP3uL41CCy3h9gcv4JmBb9OTfRuIT8yddoYRoRe/Pkh1br voz2ba2zodOxf9zqbIQujuHftrcozo1E1JWd6pEeRuVSSbnRQ2RYo5PeZpKjeRWgegOn wKaw==; dara=google.com ARC-Authentication-Results: i=1; gmr-mx.google.com; dkim=pass header.i=@gmail.com header.s=20230601 header.b=fgMHf7lE; spf=pass (google.com: domain of bnagaev@gmail.com designates 2607:f8b0:4864:20::436 as permitted sender) smtp.mailfrom=bnagaev@gmail.com; dmarc=pass (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com; dara=pass header.i=@googlegroups.com Received: from mail-pf1-x436.google.com (mail-pf1-x436.google.com. [2607:f8b0:4864:20::436]) by gmr-mx.google.com with ESMTPS id af79cd13be357-7d09a0e295dsi46826985a.2.2025.06.02.15.51.02 for (version=TLS1_3 cipher=TLS_AES_128_GCM_SHA256 bits=128/128); Mon, 02 Jun 2025 15:51:02 -0700 (PDT) Received-SPF: pass (google.com: domain of bnagaev@gmail.com designates 2607:f8b0:4864:20::436 as permitted sender) client-ip=2607:f8b0:4864:20::436; Received: by mail-pf1-x436.google.com with SMTP id d2e1a72fcca58-742c46611b6so5881893b3a.1 for ; Mon, 02 Jun 2025 15:51:02 -0700 (PDT) X-Gm-Gg: ASbGncvo90apVF63AnIdJymEBZwF+oy3YCIOjhyQ9sVsKlFWiUodSmxn9XMr/LmZMET lpWGKfMrlo733mzg4DHsXJnYIRtRBh7K2hWUHmOYllFBsMlN2n9EQJ2KgPLG+tvYNdAt45hJJHi rrD3qNs8ECUjom8LKCqhDKwN+1tw/WZAg= X-Received: by 2002:a05:6a21:339b:b0:1f5:9024:3254 with SMTP id adf61e73a8af0-21adff4c22cmr22914835637.6.1748904661138; Mon, 02 Jun 2025 15:51:01 -0700 (PDT) MIME-Version: 1.0 References: <402db6ba-2497-4aab-9f84-0d66b4b8efccn@googlegroups.com> In-Reply-To: <402db6ba-2497-4aab-9f84-0d66b4b8efccn@googlegroups.com> From: Nagaev Boris Date: Mon, 2 Jun 2025 19:50:23 -0300 X-Gm-Features: AX0GCFvJ-kqPnIKxOdXNjJpXp6KQ1NxxYNdDv_PZVMAB07QJpgP6293q2N6E0nk Message-ID: Subject: Re: [bitcoindev] Post-Quantum commit / reveal Fawkescoin variant as a soft fork To: "waxwing/ AdamISZ" Cc: Bitcoin Development Mailing List Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Original-Sender: bnagaev@gmail.com X-Original-Authentication-Results: gmr-mx.google.com; dkim=pass header.i=@gmail.com header.s=20230601 header.b=fgMHf7lE; spf=pass (google.com: domain of bnagaev@gmail.com designates 2607:f8b0:4864:20::436 as permitted sender) smtp.mailfrom=bnagaev@gmail.com; dmarc=pass (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com; dara=pass header.i=@googlegroups.com Precedence: list Mailing-list: list bitcoindev@googlegroups.com; contact bitcoindev+owners@googlegroups.com List-ID: X-Google-Group-Id: 786775582512 List-Post: , List-Help: , List-Archive: , List-Unsubscribe: , X-Spam-Score: -0.5 (/) Hi Adam, hi list, Thank you for clarifying the nuances of the scheme! > that same system but with keypath spending invalidated, so it's QR becaus= e of the hash function, but also, the tapleaf contains a QR or PQC signing = scheme in it I assume that some future QR signing algorithm might be added in a new segwit version, resembling the current Taproot structure but with keypath spending disabled (or otherwise addressed, so it doesn't present a way to steal funds in a post-quantum world), and a new opcode added to tapscript to enable QR signing. Also, such an address type would need to handle the EC opcodes if they remain enabled in tapscript. I didn't go deep into the details of this hypothetical address type, as it's a separate complex topic. For the purpose of this discussion, I assume that some version of it will be deployed before or alongside the proposed scheme. > I believe the *real* point here is not "nobody knows you committed someth= ing" though that may be practically significant, it's instead "nobody knows= the exact commitment value (hash) you used". I think this is a good observation! We can clarify it further: *nobody should be able to replicate your commitment*. I've identified a potential attack that, if it were possible, would undermine this property. Let's assume, for a second, that a new address type is just a Merkle root of a Merkle tree. Then an attacker could build their own Merkle tree on top of the original Merkle root. The attacker's tree would include the legitimate Merkle tree as a subtree. If they learn the EC pubkey and the full path in the original Merkle tree from the pubkey to the root, they could reconstruct the full path in their tree embedding the original. They could learn this information when the output is spent, so they would have everything needed to produce a malicious proof. Fortunately, with the Taproot we have today, such an extension isn't possible. There's a step (among others) involving a tagged hash function (h_tapTweak) that hashes the internal pubkey and Merkle root together before producing the final public key. This prevents constructing a Taproot output embedding another tree as a subtree based solely on the output address. We should ensure that any future taproot-like address type retains this property to prevent the attack described above. The scheme must be carefully analyzed to ensure it satisfies this crucial property: nobody should be able to replicate your commitment. I'm curious if anyone sees further edge cases we should consider. > In these analyses, I think it's common to overlook a potentially crucial = point: the UTXO set is enumerable in practical time, so we must always reme= mber that we can check our calculations against existing addresses, even if= they are hash-covered keys. I would propose hardening the requirement further: assume the attacker already knows exactly which EC address is protected with which QR output from the beginning. Even with this knowledge, the attacker should not be able to succeed. If we prove the scheme secure under these strong assumptions (which are very favorable for the attacker), then it is also secure under weaker assumptions where the attacker has less information. > This commitment is *not* literally perfectly hiding as in a properly form= ed Pedersen commitment I agree that the hiding property of a properly formed Pedersen commitment is stronger than in the proposed scheme. A Pedersen commitment is perfectly hiding: the committed value is statistically independent of the commitment, and even an attacker with unbounded computational power cannot learn anything about the value. In contrast, the proposed scheme relies on a hash of the pubkey for hiding, which is only computationally hiding: it depends on the preimage resistance of the hash function and can be weakened if quantum attacks on hash functions improve. However, I believe we cannot use Pedersen commitments here, because their binding property would be broken by a quantum computer. Pedersen commitments rely on the hardness of the discrete logarithm problem (DLP) in an elliptic curve group for binding, and Shor's algorithm can efficiently solve DLP on a quantum computer. This would allow an attacker, once the legitimate opening (v,r) is revealed, to compute an alternative opening (v',r') for the same commitment C. As a result, if Pedersen commitments were used in the scheme, they would be vulnerable during the reveal phase under a quantum attack. Best, Boris On Mon, Jun 2, 2025 at 3:41=E2=80=AFPM waxwing/ AdamISZ wrote: > > Hi Boris, list, > > > In my scheme, a user creates a QR output that > commits to a hash of a pubkey inside a Taproot leaf. This commitment > is hidden until revealed at spend time. Later, when the user wants to > spend a legacy EC output, they must spend this QR output in the same > transaction, and it must be at least X blocks old. > > There's some nuances here that seem quite interesting. If an output commi= ts (in a quantum-resistant way, i.e. hashing let's say) only to a pubkey, a= nd not a pubkey plus spending-transaction, then, in the general case, that = has the weakness that the commitment can be replicated, in another commitme= nt tx, at the time of insertion of commitment into the blockchain; so that = if the tiebreaker is "first commitment in the block(chain)" an attacker can= mess with you. In your case you refer to "a QR output that commits to the = hash of a pubkey inside a taproot leaf" but I'm finding it a tiny bit uncle= ar what you mean there. Taproot itself isn't quantum resistant (QR), so Q = =3D P + H(P,S)G is not QR even if P is NUMS. Whereas you might mean: that s= ame system but with keypath spending invalidated, so it's QR because of the= hash function, but also, the tapleaf contains a QR or PQC signing scheme i= n it (I guess this is what you actually mean). Or, you might mean something= that is structurally the same as taproot but not using secp/BIP340, but in= stead a PQC scheme with a homomorphism so that that same design can be reus= ed; call that "taproot2" .. though I don't think I've heard people talking = about that. > > Then there's what I think you focus on: the commitment is hidden. To othe= r readers, in case of confusion: I believe the *real* point here is not "no= body knows you committed something" though that may be practically signific= ant, it's instead "nobody knows the exact commitment value (hash) you used"= . > > So, I believe that's a correct/valid point that actually *doesn't* depend= on which "version" of taproot as per above. Focusing on current-taproot-bu= t-script-path-only-with-QR-in-tapleaf: we have a protection that the quant= um attacker cannot find the S in the (P,S) tuple (indeed, they cannot even = know the H in P + H(P,S)G). This commitment is *not* literally perfectly hi= ding as in a properly formed Pedersen commitment ([1]) but I do think you h= ave the normal preimage resistance against revelation as expected even in p= ost-quantum. So that prevents the "copy the commitment" problem I started o= ut by mentioning. [2] > > If "taproot2" instead then we have something for which even the keypath i= sn't crackable so I guess it's obvious. > > > Since the commitment doesn't include a txid, the user can precommit to > the pubkey hash far in advance, before knowing the details of the > eventual transaction. > > Again, I believe you're right here, but we should try to unpack what's di= fferent; because the "reveal" step of commit-reveal is accompanied by the Q= R signing event of the pre-existing QR output, we have a sane security mode= l, so there's no need to commit to the transaction in the preliminary step,= as far as I can tell. > > > More efficient use of block space > > Makes sense. > > I think the only downside I see here is that the initial commitment step = requires the PQC scheme to actually exist. That may not seem like a big dea= l, but I have a suspicion it actually will be. I think a protocol in which = we just rely on existing hash primitives and put off the PQC scheme choosin= g event may be necessary .. though I could be for sure wrong, in more than = one way, in saying that. > > Apart from that point I think this scheme seems good (as you mention, it = has the virtue of not requiring new databases etc which is pretty huge). > > Cheers, > AdamISZ/waxwing > > [1] In these analyses, I think it's common to overlook a potentially cru= cial point: the utxo set is enumerable in practical time, so we must always= remember that we can check our calculations against existing addresses, ev= en if they are hash-covered keys. > > [2] The only caveat is if you're considering the possibility of the attac= ker knowing the key in advance of even the commitment step; generally, that= 's "game over", but i know that there is some attempt to analyze that case = in some places, too. Not here. > > On Wednesday, May 28, 2025 at 6:54:21=E2=80=AFPM UTC-3 Nagaev Boris wrote= : >> >> Hi Tadge, >> >> Thanks for writing this up! The proposal is very thoughtful, and it's >> great to see concrete work on post-quantum commit/reveal schemes. >> >> I've been exploring a related approach based on a similar >> commit/reveal idea. In my scheme, a user creates a QR output that >> commits to a hash of a pubkey inside a Taproot leaf. This commitment >> is hidden until revealed at spend time. Later, when the user wants to >> spend a legacy EC output, they must spend this QR output in the same >> transaction, and it must be at least X blocks old. >> >> https://groups.google.com/g/bitcoindev/c/jr1QO95k6Uc/m/lsRHgIq_AAAJ >> >> This approach has a few potential advantages: >> >> 1. No need for nodes to track a new commitment store >> >> Because the commitment remains hidden in a Tapleaf until the spend, >> observers (including attackers) don't see it, and nodes don't need to >> store or validate any external commitment set. The only requirement is >> that the QR output must be old enough, and Bitcoin Core already tracks >> coin age, which is needed to validate existing consensus rules. >> >> 2. Commitment can be made before the transaction is known >> >> Since the commitment doesn't include a txid, the user can precommit to >> the pubkey hash far in advance, before knowing the details of the >> eventual transaction. This allows greater flexibility: you can delay >> choosing outputs, fee rates, etc., until spend time. Only knowledge of >> the EC pubkey needs to be proven when creating the QR output. >> >> 3. More efficient use of block space >> >> Multiple EC coins can be spent together with a single QR output, >> holding EC pubkey commitments in Taproot leaves. If EC coins share the >> same EC pubkey (e.g., come from the same address), they can reuse the >> same commitment. >> >> Would love to hear your thoughts on this variant. I think this one >> might be a simpler, lower-overhead option for protecting EC outputs >> post-QC. >> >> Best, >> Boris >> >> On Wed, May 28, 2025 at 2:28=E2=80=AFPM Tadge Dryja = wrote: >> > >> > One of the tricky things about securing Bitcoin against quantum comput= ers is: do you even need to? Maybe quantum computers that can break secp256= k1 keys will never exist, in which case we shouldn't waste our time. Or may= be they will exist, in not too many years, and we should spend the effort t= o secure the system against QCs. >> > >> > Since people disagree on how likely QCs are to arrive, and what the ti= ming would be if they do, it's hard to get consensus on changes to bitcoin = that disrupt the properties we use today. For example, a soft fork introduc= ing a post-quantum (PQ) signature scheme and at the same time disallowing n= ew secp256k1 based outputs would be great for strengthening Bitcoin against= an oncoming QC. But it would be awful if a QC never appears, or takes deca= des to do so, since secp256k1 is really nice. >> > >> > So it would be nice to have a way to not deal with this issue until *a= fter* the QC shows up. With commit / reveal schemes Bitcoin can keep workin= g after a QC shows up, even if we haven't defined a PQ signature scheme and= everyone's still got P2WPKH outputs. >> > >> > Most of this is similar to Tim Ruffing's proposal from a few years ago= here: >> > https://gnusha.org/pi/bitcoindev/1518710367.3...@mmci.uni-saarland.de/ >> > >> > The main difference is that this scheme doesn't use encryption, but a = smaller hash-based commitment, and describes activation as a soft fork. I'l= l define the two types of attacks, a commitment scheme, and then say how it= can be implemented in bitcoin nodes as a soft fork. >> > >> > This scheme only works for keys that are pubkey hashes (or script hash= es) with pubkeys that are unknown to the network. It works with taproot as = well, but there must be some script-path in the taproot key, as keypath spe= nds would no longer be secure. >> > >> > What to do with all the keys that are known is another issue and indep= endent of the scheme in this post (it's compatible with both burning them a= nd leaving them to be stolen) >> > >> > For these schemes, we assume there is an attacker with a QC that can c= ompute a quickly compute a private key from any secp256k1 public key. We al= so assume the attacker has some mining power or influence over miners for t= heir attacks; maybe not reliably, but they can sometimes get a few blocks i= n a row with the transactions they want. >> > >> > "Pubkey" can also be substituted with "script" for P2SH and P2WSH outp= ut types and should work about the same way (with caveats about multisig). = The equivalent for taproot outputs would be an inner key proving a script p= ath. >> > >> > ## A simple scheme to show an attack >> > >> > The simplest commit/reveal scheme would be one where after activation,= for any transaction with an EC signature in it, that transaction's txid mu= st appear in a earlier transaction's OP_RETURN output. >> > >> > When a user wants to spend their coins, they first sign a transaction = as they would normally, compute the txid, get that txid into an OP_RETURN o= utput somehow (paying a miner out of band, etc), then after waiting a while= , broadcast the transaction. Nodes would check that the txid matches a prev= iously seen commitment, and allow the transaction. >> > >> > One problem with this scheme is that upon seeing the full transaction,= the attacker can compute the user's private key, and create a new commitme= nt with a different txid for a transaction where the attacker gets all the = coins. If the attacker can get their commitment and spending transaction in= before the user's transaction, they can steal the coins. >> > >> > In order to mitigate this problem, a minimum delay can be enforced by = consensus. A minimum delay of 100 blocks would mean that the attacker would= have to prevent the user's transaction from being confirmed for 100 blocks= after it showed up in the attacker's mempool. The tradeoff is that longer = periods give better safety at the cost of more delay in spending. >> > >> > This scheme, while problematic, is better than nothing! But it's possi= ble to remove this timing tradeoff. >> > >> > >> > ## A slightly more complex scheme with (worse) problems >> > >> > If instead of just the txid, the commitment were both the outpoint bei= ng spent, and the txid that was going to spend it, we could add a "first se= en" consensus rule. Only the first commitment pointing to an outpoint works= . >> > >> > So if nodes see two OP_RETURN commitments in their sequence of confirm= ed transactions: >> > >> > C1 =3D outpoint1, txid1 >> > C2 =3D outpoint1, txid2 >> > >> > They can ignore C2; C1 has already laid claim to outpoint1, and the tr= ansaction identified by txid1 is the only transaction that can spend outpoi= nt1. >> > >> > If the user manages to get C1 confirmed first, this is great, and elim= inates the timing problem in the txid only scheme. But this introduces a di= fferent problem, where an attacker -- in this case any attacker, even one w= ithout a QC -- who can observe C1 before it is confirmed can flip some bits= in the txid field, freezing the outpoint forever. >> > >> > We want to retain the "first seen" rule, but we want to also be able t= o discard invalid commitments. In a bit flipping attack, we could say an in= valid commitment is one where there is no transaction described by the txid= . A more general way to classify a commitment as invalid is a commitment ma= de without knowledge of the (secret) pubkey. Knowledge of the pubkey is wha= t security of coins is now hinging on. >> > >> > >> > The actual commitment scheme >> > >> > >> > We define some hash function h(). We'll use SHA256 for the hashing, bu= t it needs to be keyed with some tag, for example "Alas poor Koblitz curve,= we knew it well". >> > >> > Thus h(pubkey) is not equal to the pubkey hash already used in the bit= coin output script, which instead is RIPEMD160(SHA256(pubkey)), or in bitco= in terms, HASH160(pubkey). Due to the hash functions being different, A =3D= HASH160(pubkey) and B =3D h(pubkey) will be completely different, and nobo= dy should be able to determine if A and B are hashes of the same pubkey wit= hout knowing pubkey itself. >> > >> > An efficient commitment is: >> > >> > C =3D h(pubkey), h(pubkey, txid), txid >> > (to label things: C =3D AID, SDP, CTXID) >> > >> > This commitment includes 3 elements: a different hash of the pubkey wh= ich will be signed for, a proof of knowledge of the pubkey which commits to= a transaction, and an the txid of the spending transaction. We'll call the= se "address ID" (AID), sequence dependent proof (SDP), and the commitment t= xid (CTXID). >> > >> > For those familiar with the proposal by Ruffing, the SDP has a similar= function to the authenticated encryption part of the encrypted commitment.= Instead of using authenticated encryption, we can instead just use an HMAC= -style authentication alone, since the other data, the CTXID, is provided. >> > >> > When the user's wallet creates a transaction, they can feed that trans= action into a commitment generator function which takes in a transaction, e= xtracts the pubkey from the tx, computes the 3 hashes, and returns the 3-ha= sh commitment. Once this commitment is confirmed, the user broadcasts the t= ransaction. >> > >> > Nodes verify the commitment by using the same commitment generator fun= ction and checking if it matches the first valid commitment for that AID, i= n which case the tx is confirmed. >> > >> > If a node sees multiple commitments all claiming the same AID, it must= store all of them. Once the AID's pubkey is known, the node can distinguis= h which commitments are valid, which are invalid, and which is the first se= en valid commitment. Given the pubkey, nodes can determine commitments to b= e invalid by checking if SDP =3D h(pubkey, CTXID). >> > >> > As an example, consider a sequence of 3 commitments: >> > >> > C1 =3D h(pubkey), h(pubkey', txid1), txid1 >> > C2 =3D h(pubkey), h(pubkey, txid2), txid2 >> > C3 =3D h(pubkey), h(pubkey, txid3), txid3 >> > >> > The user first creates tx2 and tries to commit C2. But an attacker cre= ates C1, committing to a different txid where they control the outputs, and= confirms it first. This attacker may know the outpoint being spent, and ma= y be able to create a transaction and txid that could work. But they don't = know the pubkey, so while they can copy the AID hash, they have to make som= ething up for the SDP. >> > >> > The user gets C2 confirmed after C1. They then reveal tx2 in the mempo= ol, but before it can be confirmed, the attacker gets C3 confirmed. C3 is a= valid commitment made with knowledge of the pubkey. >> > >> > Nodes can reject transactions tx1 and tx3. For tx1, they will see that= the SDP doesn't match the data in the transaction, so it's an invalid comm= itment. For tx3, they will see that it is valid, but by seeing tx3 they wil= l also be able to determine that C2 is a valid commitment (since pubkey is = revealed in tx3) which came prior to C3, making C2 the only valid commitmen= t for that AID. >> > >> > >> > ## Implementation >> > >> > Nodes would keep a new key/value store, similar to the existing UTXO s= et. The indexing key would be the AID, and the value would be the set of al= l (SDP, CTXID) pairs seen alongside that AID. Every time an commitment is s= een in an OP_RETURN, nodes store the commitment. >> > >> > When a transaction is seen, nodes observe the pubkey used in the trans= action, and look up if it matches an AID they have stored. If not, the tran= saction is dropped. If the AID does match, the node can now "clean out" an = AID entry, eliminating all but the first valid commitment, and marking that= AID as final. If the txid seen matches the remaining commitment, the trans= action is valid; if not, the transaction is dropped. >> > >> > After the transaction is confirmed the AID entry can be deleted. Delet= ing the entries frees up space, and would allow another round to happen wit= h the same pubkey, which would lead to theft. Retaining the entries takes u= p more space on nodes that can't be pruned, and causes pubkey reuse to dest= roy coins rather than allow them to be stolen. That's a tradeoff, and I per= sonally guess it's probably not worth retaining that data but don't have a = strong opinion either way. >> > >> > Short commitments: >> > >> > Since we're not trying to defend against collision attacks, I think al= l 3 hashes can be truncated to 16 bytes. The whole commitment could be 48 b= ytes long. Without truncation the commitments would be 96 bytes. >> > >> > >> > ## Activation >> > >> > The activation for the commit/reveal requirement can be triggered by a= proof of quantum computer (PoQC). >> > >> > A transaction which successfully spends an output using tapscript: >> > >> > OP_SHA256 OP_CHECKSIG >> > >> > is a PoQC in the form of a valid bitcoin transaction. In order to sati= sfy this script, the spending transaction needs to provide 2 data elements:= a signature, and some data that when hashed results in a pubkey for which = that signature is valid. If such a pair of data elements exists, it means t= hat either SHA256 preimage resistance is broken (which we're assuming isn't= the case) or someone can create valid signatures for arbitrary elliptic cu= rve points, ie a cryptographically relevant quantum computer (or any other = process which breaks the security of secp256k1 signatures) >> > >> > Once such a PoQC has been observed in a confirmed transaction, the req= uirements for the 3-hash commitment scheme can be enforced. This is a soft = fork since the transactions themselves look the same, the only requirement = is that some OP_RETURN outputs show up earlier. Nodes which are not aware o= f the commitment requirement will still accept all transactions with the ne= w rules. >> > >> > Wallets not aware of the new rules, however, are very dangerous, as th= ey may try to broadcast signed transactions without any commitment. Nodes t= hat see such a transaction should drop the tx, and if possible tell the wal= let that they are doing something which is now very dangerous! On the open = p2p network this is not really enforceable, but people submitting transacti= ons to their own node (eg via RPC) can at least get a scary error message. >> > >> > >> > ## Issues >> > >> > My hope is that this scheme would give some peace of mind to people ho= lding bitcoin, that in the face of a sudden QC, even with minimal preparati= on their coins can be safe at rest and safely moved. It also suggests some = best practices for users and wallets to adopt, before any software changes:= Don't reuse addresses, and if you have taproot outputs, include some kind = of script path in the outer key. >> > >> > There are still a number of problems, though! >> > >> > - Reorgs can steal coins. An attacker that observes a pubkey and can r= eorg back to before the commitment can compute the private key, sign a new = transaction and get their commitment in first on the new chain. This seems = unavoidable with commit/reveal schemes, and it's up to the user how long th= ey wait between confirming the commitment and revealing the transaction. >> > >> > - How to get op_returns in >> > If there are no PQ signature schemes activated in bitcoin when this ac= tivates, there's only one type of transaction that can reliably get the OP_= RETURN outputs confirmed: coinbase transactions. Getting commitments to the= miners and paying them out of band is not great, but is possible and we se= e this kind of activity today. Users wouldn't need to directly contact mine= rs: anyone could aggregate commitments, create a large transaction with man= y OP_RETURN outputs, and then get a miner to commit to that parent transact= ion. Users don't need to worry about committing twice as identical commitme= nts would be a no op. >> > >> > - Spam >> > Anyone can make lots of OP_RETURN commitments which are just random nu= mbers, forcing nodes to store these commitments in a database. That's not g= reat, but isn't much different from how bitcoin works today. If it's really= a problem, nodes could requiring the commitment outputs to have a non-0 am= ount of bitcoin, imposing a higher cost for the commitments than other OP_R= ETURN outputs. >> > >> > - Multiple inputs >> > If users have received more than one UTXO to the same address, they wi= ll need to spend all the UTXOs at once. The commitment scheme can deal with= only the first pubkey seen in the serialized transaction. >> > >> > - Multisig and Lightning Network >> > If your multisig counterparties have a QC, multisig outputs become 1 o= f N. Possibly a more complex commit / reveal scheme could deal with multipl= e keys, but the keys would all have to be hashed with counterparties not kn= owing each others' unhashed pubkeys. This isn't how existing multisig outpu= ts work, and in fact the current trend is the opposite with things like Mus= ig2, FROST and ROAST. If we're going to need to make new signing software a= nd new output types it might make more sense to go for a PQ signature schem= e. >> > >> > - Making more p2wpkhs >> > You don't have to send to a PQ address type with these transactions --= you can send to p2wpkh and do the whole commit/reveal process again when y= ou want to spend. This could be helpful if PQ signature schemes are still b= eing worked on, or if the PQ schemes are more costly to verify and have hig= h fees in comparison to the old p2wpkh output types. It's possible that in = such a scenario a few high-cost PQ transactions commit to many smaller EC t= ransactions. If this actually gets adoption though, we might as well drop t= he EC signatures and just make output scripts into raw hash / preimage pair= s. It could make sense to cover some non-EC script types with the same 3-ha= sh commitment requirement to enable this. >> > >> > ## Conclusion >> > >> > This PQ commit / reveal scheme has similar properties to Tim Ruffing's= , with a smaller commitment that can be done as a soft fork. I hope somethi= ng like this could be soft forked with a PoQC activation trigger, so that i= f a QC never shows up, none of this code gets executed. And people who take= a couple easy steps like not reusing addresses (which they should anyway f= or privacy reasons) don't have to worry about their coins. >> > >> > Some of these ideas may have been posted before; I know of the Fawksco= in paper (https://jbonneau.com/doc/BM14-SPW-fawkescoin.pdf) and the recent = discussion which linked to Ruffing's proposal. Here I've tried to show how = it could be done in a soft fork which doesn't look too bad to implement. >> > >> > I've also heard of some more complex schemes involving zero knowledge = proofs, proving things like BIP32 derivations, but I think this gives some = pretty good properties without needing anything other than good old SHA256. >> > >> > Hope this is useful & wonder if people think something like this would= be a good idea. >> > >> > -Tadge >> > >> > -- >> > You received this message because you are subscribed to the Google Gro= ups "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/bitcoi= ndev/cc2f8908-f6fa-45aa-93d7-6f926f9ba627n%40googlegroups.com. >> >> >> >> -- >> Best regards, >> Boris Nagaev > > -- > 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/bitcoinde= v/402db6ba-2497-4aab-9f84-0d66b4b8efccn%40googlegroups.com. --=20 Best regards, Boris Nagaev --=20 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 e= mail to bitcoindev+unsubscribe@googlegroups.com. To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/= CAFC_Vt6t9QvjUVJ_N2kYh60iiB3MgPkrahQ97CoTQSPFqdQ3yg%40mail.gmail.com.