From mboxrd@z Thu Jan 1 00:00:00 1970 Delivery-date: Wed, 13 Nov 2024 14:23:37 -0800 Received: from mail-qt1-f186.google.com ([209.85.160.186]) by mail.fairlystable.org with esmtps (TLS1.3) tls TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256 (Exim 4.94.2) (envelope-from ) id 1tBLlw-0000vQ-8d for bitcoindev@gnusha.org; Wed, 13 Nov 2024 14:23:37 -0800 Received: by mail-qt1-f186.google.com with SMTP id d75a77b69052e-460a85907b7sf110359871cf.0 for ; Wed, 13 Nov 2024 14:23:35 -0800 (PST) ARC-Seal: i=2; a=rsa-sha256; t=1731536610; cv=pass; d=google.com; s=arc-20240605; b=bALCS13VM/rNQFhRh6ZVb8zGcq7ETdPZKNDvHHC7PgCaXdtQa2SHG1j7BoDL0NVHOA I7vA9n0kd9P3CisqJDWph7tMkDWoO7ac1TmeHfToYqO2SsYuc4J4Y8dHHth7299ASYKF f7+2MVzG71HGFAMW9Xj9jeHJR+klXU1wL44/fWDFbSbzs36MUJOcGsEm6L1i9hIfU4cS GYcScvatu1S6Si2V/1R0LT2WFW6WduxDKeNvs30s0zClWpv3IBfOnaTELPFFgLsa8S/1 M0ZdOnXuJeKBlmmMJRwBk2cj3EDEDeiRdv+DWn8nZ/7bf+VlrmyxmyikqNmdi7U2wnYd cGEg== 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=ZAQFQDUvxRY/1+5iXAkeNBzma5q5Qu0drCvpzmi4ows=; fh=55qd1FlaY8vmW4jwqZPLM8ivEcPSD3Tbka5fJ+vIc7Y=; b=VmP7bM5bY1Y57LhjxpzMzDo1f6itA6orhUL63zyOET62o7GggQpCd2V3IOWD1Ul5Nn PVh2hmB3r7yAmBLbDB1Rzg1SubEBXkEWEa9RcfxhRR7Ha5y5W1cLSULD1H+6UyiUu220 SIi8tEkw8tZsTM9AYTBS0b/gJ5VCQAgdghVlsNN2lgz0QKgOKUdzZlZN/QiPNMApLsgz aFFqoqSdVkKxlkMcJ1f1PZDnUzX5EJANnbV4qR9KBTVh8GF6/uHkRfZOnv9ae4rCM4TU xGHkeku8VFeexXZS+wFHDkuV6OxNAkzHTl9ErqCfeMa3D5+kdWbyWVPyKpzdpC4ywP7a pqPQ==; darn=gnusha.org ARC-Authentication-Results: i=2; gmr-mx.google.com; dkim=pass header.i=@gmail.com header.s=20230601 header.b=f7xgozrl; spf=pass (google.com: domain of eth3rs@gmail.com designates 2a00:1450:4864:20::62d as permitted sender) smtp.mailfrom=eth3rs@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=1731536610; x=1732141410; 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=ZAQFQDUvxRY/1+5iXAkeNBzma5q5Qu0drCvpzmi4ows=; b=QYU4USqhKTQp96kcHizAzoDjAlbjwileezdAlYcINMcLG3sDeXisG0K5dW9s8HD8Ia f9P4RxZ37sULIdMkkbMWkkYDvTu+u2VeIFJt5vxtazGvhaPvZqGShnMrwe2lCvTyCfcZ vjhAymsyVlw+zcfvxgxsIfJ+VOoCmOiu4+5VwFsBxCaOGCHBQdIyPiH2shZMjVieIx9C cD+lX+n1nJjMhXToWE7VvMCaDPRXPfd9j6wjwBl9DQUbWNiPDYIF2YpMF6VyC0AvxDN3 D0+9ejrZmGYN8wbpgg/1d2ZtcF2/R/wPLunzn4P6f+vcfW4h8uDG1EuUKnt3c4ycZmU4 upug== DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20230601; t=1731536610; x=1732141410; 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=ZAQFQDUvxRY/1+5iXAkeNBzma5q5Qu0drCvpzmi4ows=; b=cL4b+NPUsIjBiFSrNrA1iOiqBNdGCd1m8fvqhJsWnhqy0/ET0mJiS4yzEFUtlN6Mno maskmxZj93fr1hNZf4Sw7Jbiqsz9sJTrT7C5j5yX+nGSqE2wrklHPCGMutMmrqkTikze guM1xYcz1syNw7AHbh6ReNS00uVqfsKj5deGDfte38XxcJDxVCbKQlYh/9jsQbaLLY5F l9EJZPhseGWYMc75gu7qx40AW4ojeyYGT8RGqeQ7nYLLWvtY1b/UtxRem4+1r6TWlldW 8IuK/SJ5rLlVHvnZJcsscSmJznHxOR7lnDzmb3vCgR551z7EV9Kh2hv0TI4hNel+NBQp 5aHA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1731536610; x=1732141410; 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=ZAQFQDUvxRY/1+5iXAkeNBzma5q5Qu0drCvpzmi4ows=; b=B2KNeAYQguaBAI3725UOMgiqp2kc7eorCUqQttJDAo3ipDUYGV18eUVLJ5nAJ9LQe9 1A6Xmn7Gu8dVk8jGVWcfFfFaVdL3lbThMWQOawKmm4HyanqVh8oP89l8GDwbiUHX9cRs Wh8g/RduuRedizyzHbwna2AmP9RO8Qpzo/fsnJdqmFIl+1zboKGpzUThyf3DKvpUkpCc mp5buqvLY1g/6uFAa74UwUN1mdJ2B3QRc8cD0lA470PZgI/s3kYl2xsT98llacFukrig o670yDVhH0BOAyaae4PnQEs1EezBiksnlDDvJIlXxMkaNZ+B0nVYFuMuRu2N68Q4Xt2S w52A== Sender: bitcoindev@googlegroups.com X-Forwarded-Encrypted: i=2; AJvYcCUXuMCDm9rfjGToMUqPaFt9aTGboaer+0uleTqBcMbNQRntEXZaDzVsEbGEdwzKZ2wXW4yKhGxwl6c1@gnusha.org X-Gm-Message-State: AOJu0YyPf+QedezEKX8m6xDLWS0eACpzGAYbEAVrVpWbW/t/sr+4SyMZ lzo/Dq3LqowKjcDktfITYVxPYY/23M3hYSjqbSkY00C/LenpXQ2+ X-Google-Smtp-Source: AGHT+IG1jtY8CX9SZm/MlJ2Tq9Zds3FcE1RcEPjLK7hXAtN3yZ668Y3Qct1KC0l7ytnpFQaUYv50Xw== X-Received: by 2002:ac8:7fc3:0:b0:462:d75a:e2f0 with SMTP id d75a77b69052e-46309319448mr244701911cf.7.1731536609463; Wed, 13 Nov 2024 14:23:29 -0800 (PST) X-BeenThere: bitcoindev@googlegroups.com Received: by 2002:a05:622a:410c:b0:460:e9a4:81fc with SMTP id d75a77b69052e-46356f8a767ls1921661cf.2.-pod-prod-06-us; Wed, 13 Nov 2024 14:23:26 -0800 (PST) X-Received: by 2002:a05:620a:f0e:b0:7b1:3fe8:8f11 with SMTP id af79cd13be357-7b331eac97fmr2948496985a.33.1731536606751; Wed, 13 Nov 2024 14:23:26 -0800 (PST) Received: by 2002:a05:620a:2284:b0:7a1:c409:aa2c with SMTP id af79cd13be357-7b35b0a1b52ms85a; Wed, 13 Nov 2024 14:07:06 -0800 (PST) X-Received: by 2002:a05:600c:1c1d:b0:431:5957:27e8 with SMTP id 5b1f17b1804b1-432d4ad89a3mr34169965e9.28.1731535624421; Wed, 13 Nov 2024 14:07:04 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1731535624; cv=none; d=google.com; s=arc-20240605; b=MpA8HkdedDEgoeUIVeSjzfJSyDGLnyvQfQ3PwqxQB8XTRAep07KqgeUn4kITJeoOk4 G0mHZ/+nGG75GzdDvqZbHK/FfFeeL6taT0SY08fOrDaw5oeP4WPgyWEywdpU9yBvB7RF HDbH4Ka0Gf+Qgfd4umpTNQdkd6JylvBDaLoVS6OYw1lpw2l68XwJxyKQvTq/7WpeH8js 7OV16YatvplaxEeEQYEfjtVLDmHzK+uh0IsbHx7+D16rUZsj/g7FI0ESWVWKE4A2BcC9 2GTkWaeQffsvneLtRVk1DhJnrisC5kXXmhM2OUBuYt0Uza8A2XX2hUrPYEPlidqACQHa vcKw== 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=+0LAo2cWgyG+mazqAgjsDK+kWqzBv8uNB5cFVrHjngE=; fh=sapDHqhE46zLmMBeB1lkoe0zq8J9+V3Afx71/j8kvug=; b=BahH0bnIjoEh0X04YcWyJemtvQ/pR1IC0iRoCpVGM1bMeWhHD7YdRIi8m42f0lhTW0 2igpZXxDZdJehQ8h+yc99Musol8OqClDWSYm9/gohHCT3YhC7DMSj3Zjaitv/I6Gfu/d nS+Eon21I/0Fd1icAxqYGt4EaFl83KGs3N/Uw8yoUCRPlGuzEQgxjzz0oNhYmSQJWiHm FCuXjbKFObrTyqyoQVfpMsJZXMqCehK6lf8fl1MYw0IBSwuJbXmOr2rAk7Cx6yDnVgYl mWKQdME6N+foPvDSsHjR4fadEQfZn+W0cABujxR2A2Q2zZUYjp7LIKBNq9psuRmUD8tD jV/Q==; dara=google.com ARC-Authentication-Results: i=1; gmr-mx.google.com; dkim=pass header.i=@gmail.com header.s=20230601 header.b=f7xgozrl; spf=pass (google.com: domain of eth3rs@gmail.com designates 2a00:1450:4864:20::62d as permitted sender) smtp.mailfrom=eth3rs@gmail.com; dmarc=pass (p=NONE sp=QUARANTINE dis=NONE) header.from=gmail.com; dara=pass header.i=@googlegroups.com Received: from mail-ej1-x62d.google.com (mail-ej1-x62d.google.com. [2a00:1450:4864:20::62d]) by gmr-mx.google.com with ESMTPS id 5b1f17b1804b1-432d478830csi1552045e9.0.2024.11.13.14.07.04 for (version=TLS1_3 cipher=TLS_AES_128_GCM_SHA256 bits=128/128); Wed, 13 Nov 2024 14:07:04 -0800 (PST) Received-SPF: pass (google.com: domain of eth3rs@gmail.com designates 2a00:1450:4864:20::62d as permitted sender) client-ip=2a00:1450:4864:20::62d; Received: by mail-ej1-x62d.google.com with SMTP id a640c23a62f3a-a9ec267b879so1310634766b.2 for ; Wed, 13 Nov 2024 14:07:04 -0800 (PST) X-Received: by 2002:a17:906:6a10:b0:a9a:bbcc:5092 with SMTP id a640c23a62f3a-aa1f810639emr442113566b.39.1731535623595; Wed, 13 Nov 2024 14:07:03 -0800 (PST) MIME-Version: 1.0 References: In-Reply-To: From: Ethan Heilman Date: Wed, 13 Nov 2024 17:06:27 -0500 Message-ID: Subject: Re: [bitcoindev] Re: ColliderScript: Covenants in Bitcoin via 160-bit hash collisions To: Antoine Riard Cc: Bitcoin Development Mailing List Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Original-Sender: eth3rs@gmail.com X-Original-Authentication-Results: gmr-mx.google.com; dkim=pass header.i=@gmail.com header.s=20230601 header.b=f7xgozrl; spf=pass (google.com: domain of eth3rs@gmail.com designates 2a00:1450:4864:20::62d as permitted sender) smtp.mailfrom=eth3rs@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 (/) > If I'm understanding correctly, the goal of the equivalence check is to = find a `y` such that `y <- h_big_script(s1)` and `y <- h_small_script(s2)` are logically equal. Once such `y` is found, the data-carrying transaction is grinded until s1 and s2 are equal. You have the basic idea correct but I think you might be missing one piece of this (or perhaps you just simplified some details). For an honest party spending a covenant the s1 and s2 they generate are always equal. This means that y1 <- h_big_script(s1) and y2 <- h_small_script(s2) will generate the same y (y1 =3D y2). As you point out we can't compare y1 to y2 because they are encoded differently. > The hash `y` outcome for both `h_big_script` and `h_small_script` will b= e never compared themselves during the script execution, as we *cannot*. Th= e former is a plain data push and the remainder an array of 32-bits script = elements. To show equivalence between s1 and s2, we find a value d <-- dGen(w,t) which is equal to y. y1 <- h_big_script(s1) d1 <-- dGen_big_script(w, t) y1 =3D=3D d1? y2 <- h_small_script(s2) d2 <-- dGen_small_script(w, t) y2 =3D=3D d2? Since the inputs to dGen are 32-bit elements in both dGen_big_script and dGen_small_script, we know just DUP w and t for evaluation in small script and big script. Since dGen is deterministic it must be the case that dGen_big_script(w, t) always equals dGen_small_script(w, t). The trick is finding a w, t, s where dGen(w, t) =3D h(s) where s =3D s1 =3D s2. > 1). could a 160-bit hash collision attacker leverage some staticness of f= ields in bip341 sighash to lower the hardness under 2^109 ? As analyzed in = appendix G. If I understand what you mean by staticness, the answer is no. Staticness should not provide any advantage to an attacker. In fact if the attacker does not sufficiently randomize the sighash on each query, they will have more difficulty, not less, finding collisions. > 2) The Schnorr trick assumes a signature where the pubkey and nonce are e= quivalent to the generator point. I think there could be some forgery of th= e covenanted transaction data, if an adversary can construct a transaction = with still the same s1 and s2 to satisfy Big Script and Small Script. I'm not sure what you mean here. Can you provide a concrete example of this attack? Thanks, Ethan On Tue, Nov 12, 2024 at 12:41=E2=80=AFPM Antoine Riard wrote: > > Hi Ethan, > > Thanks you for this astute paper. > > The crux of the paper relies on the equivalence check, which > in my understanding can be described as the following (correct > me if the set of algorithms differs). On one side, we have our > old good correct signatures on the stack. A signature is a > commitment to the signature hash fields. This signature can be > verified to be valid and it can be given to one of the 160-bits > hash functions, i.e OP_SHA1 or OP_RIPEMD160. > > E.g: < > > > Doing the Schnorr trick, a data-carrying transaction can be > altered until its signature is equivalent to the SchnorrHash, > by selecting accordingly that pubkey P and nonce R are equal > to the generator. Those elements that the signature is well- > composed are further checked by the small script "signature > defragmentation" 32-bits integers opcodes. > > On the other side, we have the 32-bits integers opcodes > that can be used to re-implement "bitcoin script natively-ish" > cryptographic operations, e.g blake3. Giving the full script > would be too lengthy, though hash functions are just (very) > smart sequences of XORed seed, key, data that one can simulate > easy in bitcoin script. E.g to flip one bit of data. > > E.g: <<1-bit input_a> <0x1> <0x0> > <0x1> > > > So let's say we have some basic cryptographic operation > like OP_SHA1 in small script (for p2tr tapscript spend > the script size limit is the block size). If I'm understanding > correctly, the goal of the equivalence check is to find a `y` > such that `y <- h_big_script(s1)` and `y <- h_small_script(s2)` > are logically equal. Once such `y` is found, the data-carrying > transaction is grinded until s1 and s2 are equal. The hash `y` > outcome for both `h_big_script` and `h_small_script` will be > never compared themselves during the script execution, as > we *cannot*. The former is a plain data push and the remainder > an array of 32-bits script elements. > > Once the equivalence check has been done, the restrictions > on the signature construction from the Schnorr trick can > be checked, and then what covenant checks can be done in > the remainder of the 4MB weight unit can be play out. E.g > checking the spending transaction nAmount twice 32 bits > are less than a given value. > > <<32-bits> > <32-bits> > > > Withstanding the code proof that a OP_SHA1 or OP_RIPEMD160 > fips-180-1 implem in small script effectively fit in the > block size, I think the colliderscript equivalent check > construction as presented can be sound, though I have > few questions on the security model. > > 1) Let's say you have a 2 step smart contract as presented > in Figure 4. The locking script is committed in a tapscript > somewhere in the tree, and as such the collision `y` to found > cannot be observed ahead of the spending. > > Once the script is revealed and before the transaction confirms > every 10 block _in average_, an adversary with enormous > computational resources, could come to find another collision > (what I think you call a triple collision). While the s1 value, > i.e the main data fields of the data-carrying transaction > shouldn't be know ahead, for some use-cases they might very > standards. > > E.g if you take a vault protocol, only the spent outpoint and > pubkey might differs among vault instance belonging to different > users (e.g everyone use same emergency or revaulting timelocks by > default). So could a 160-bit hash collision attacker leverage > some staticness of fields in bip341 sighash to lower the > hardness under 2^109 ? As analyzed in appendix G. > > 2) The Schnorr trick assumes a signature where the pubkey and > nonce are equivalent to the generator point. I think there > could be some forgery of the covenanted transaction data, if > an adversary can construct a transaction with still the same > s1 and s2 to satisfy Big Script and Small Script. > > However, the transaction data would be a security downgrade > from the smart contract protocol. E.g a nLocktime committing > to a sooner chain tip than now, or even worst a malleated > output pubkey. I think it's easy to fix if the schnorr pubkey, > nonce are committed in the covenant locking script, and explicitly > verified as such in the small script. This is pointed out > in the section 2.4 about transaction grinding, but it's not > discussed further afaict e.g in the section 7 on the security > discussion. > > Best, > Antoine > ots hash: b47373e430c25a96e642ddad3cc330ee364f06c06f81a63238926a9ebbcb679= 5 > Le jeudi 7 novembre 2024 =C3=A0 17:48:03 UTC, Ethan Heilman a =C3=A9crit = : >> >> We wanted to make bitcoin-dev aware of our recently published draft on >> how to create and spend covenants on Bitcoin using Tapscript. >> https://colliderscript.co/colliderscript.pdf >> >> Our approach does not require soft forks and should work on Bitcoin as >> it is currently deployed. While creating these covenants is as easy as >> creating a transaction with P2WSH output, spending these covenants >> requires substantial computation and involves creating very large >> bitcoin transactions. >> >> Spending such a covenant requires ~2^86 hash calls to SHA-1 and >> RIPEMD-160. In comparison, mining a Bitcoin block at current >> difficulty requires ~2^78.3 hash calls to SHA256x2. Thus, spending >> such a covenant would require the same number of hash queries the >> entire Bitcoin network does in roughly ~33 hours. Such covenants could >> be created today, but spending them likely requires the creation of >> dedicated ASICs. >> >> While the computational costs limit the immediate applicability of our >> covenants, we are optimistic that future work can significantly >> improve these numbers. This approach is not a replacement for a >> covenant opcode because: >> 1. High computational cost: Our approach is likely to be many orders >> of magnitude more computationally expensive than a covenant opcode. >> 2. 4Mb transactions: Transaction size, computation cost trade-off >> exists that reduces the computational costs of these covenants by >> increasing the transaction size. Due to most of the cost being >> computational it is likely they always be just under 4MB in size even >> with efficiency gain. 4MB is the limit for valid transactions. >> >> Our approach is framed around covenants, but our approach enables >> arbitrary computation on Bitcoin transaction data. This arbitrary >> computation is bounded only by the circuit size we can pack into what >> is left of a 4Mb Bitcoin transaction after it contains all our >> necessary machinery. This means, for example, we can do Tapscript >> lamport signatures that sign the sighash of the spending transaction. >> >> One of the authors of our paper, Andrew Poelstra, proposes in the >> paper a very interesting use case for lamport signatures (Section 7.2) >> which I think is worth making the list aware of. Leveraging the fact >> that it is very cheap for users to write and deploy our covenants, it >> is only expensive to spend them, the following scheme is proposed. A >> user could create a covenant in a tapleaf whose spending condition >> requires a Lamport signature over spending transaction. While the >> resulting script will be very large, they can hide it in a Taproot >> leaf where it will never be serialized on-chain unless it is used. In >> the case of a =E2=80=9Csurprise quantum computer=E2=80=9D which forces B= itcoin to >> suddenly disable all elliptic curve cryptography including taproot key >> spends, such users will still be able to spend their coins (though at >> enormous cost). If a large quantity of users do this, it may be >> possible for the Bitcoin chain to survive such an event, even if no >> better solution is found or deployed. >> >> >> Our Technique >> =3D=3D=3D=3D >> >> We will now look at how our technique works by introducing the core >> problem we solve and then showing how by solving this problem we can >> enforce covenants. >> >> >> Let=E2=80=99s say you want to write a Bitcoin script that checks if >> 12345678abcdef00 and [12345678, abcdef00] are equivalent. That is, if >> you treated 12345678 and abcdef00as a single element would it be >> equal to 12345678abcdef00? >> >> If we had OP_CAT this would be easy: >> 12345678abcdef00 =3D=3D CAT(12345678, abcdef00) >> >> We call checking if one element is the concatenation of a list of >> smaller elements, an equivalence check. We can ask is 12345678abcdef00 >> equivalent to [12345678, abcdef00]? >> In Bitcoin script, checking equivalence between a single big element >> and a list of small elements is quite challenging, below we will show >> how to do it. >> >> Before getting there we need to make a short digression first. It has >> been part of the lore for some time that Bitcoin script can perform >> arbitrary computation on inputs, so long as the inputs are encoded as >> a list of 32-bit stack elements. This uses opcodes like OP_ADD, >> OP_SUB, which only accept 32-bit inputs. We call functions written >> using 32-bit elements Small Script. People have built all sorts of >> things in Small Script, for instance you can compute Blake3 in >> Tapscript in a bitcoin output using only 45,000 opcodes (a.k.a. 45,000 >> bytes)! See https://bitvmx.org/knowledge/optimizing-algorithms-for-bitco= in-script >> >> Let=E2=80=99s say you have a Small Script implementation of SHA-1 where = it >> treats [12345678, ABCDEF00] as an Small Script encoding of >> 12345678abcdef00. Does the following equivalence check work? >> >> OP_SHA1(12345678abcdef00) =3D=3D smallscript_SHA1([12345678, ABCDEF00]) >> >> No, because OP_SHA1 will produce one big 160-bit (20 byte) stack element >> OP_SHA1(12345678abcdef00) =E2=86=92 a12d9ee23d07317c2d2d6887fe955819bc2d= 24c5 >> whereas the Small Script implementation of SHA1 will produce 5 32-bit >> (4 Byte) stack elements >> smallscript_SHA1([12345678, abcdef00]) =E2=86=92 [a12d9ee2, 3d07317c, >> 2d2d6887, fe955819, bc2d24c5] >> >> Checking if these are the same requires the equivalence check, the >> very thing we were trying to build. Ok, hear me out, what if you >> magically discovered a value X which was 32-bits in size and just so >> happened to collide with 12345678abcdef00. That is, >> SHA1(12345678abcdef00) =3D=3D SHA1(X) is true >> AND >> 12345678abcdef00 !=3D X >> AND >> Size(X) =3D 32-bits >> >> You could do an equivalence check for 12345678abcdef00 by doing the foll= owing: >> >> In Big Script >> Check OP_SHA1(12345678abcdef00) =3D=3D OP_SHA1(X) >> >> In Small Script >> Check smallscript_SHA1([12345678, abcdef00]) =3D=3D smallscript_SHA1(X) >> >> If both of these return true, then we decide 12345678abcdef00 is >> equivalent to [12345678, abcdef00]. >> >> Put more generally: >> Check OP_SHA1(A) =3D=3D OP_SHA1(X) >> and >> >> Check smallscript_SHA1(B) =3D=3D smallscript_SHA1(X) >> >> Now if A is equivalent to B, and X collides with A, then X will >> collide with B in Small Script because B is just the Small Script >> encoding of A. However if A is not equivalent to B then this implies a >> triple collision which is much more expensive to find: >> >> SHA1(A) =3D SHA1(X) =3D SHA1(B) >> >> where A !=3D B, A!=3DX, B!=3DX >> >> Given the choice between two possibles for an A and B that pass the chec= k: >> 1. A equivalent to B >> 2. A not actually equivalent to B requires a triple collision that is >> computationally infeasible >> We argue that 1 is much more likely. >> >> Ok, but as some of the cryptographers are now typing, if X is only >> 32-bits it is extremely unlikely to find an X that collides for any >> particular value. To exploit the birthday bound to find collisions >> with a 160-bit hash function you need the input to be >80-bit in size. >> Overcoming this problem is the last trick of the paper. We introduce >> the function dGen(w, t) which can take any length of input and that >> input can be read by Small Script and Big Script. >> >> dGen(w, t): >> y =3D SHA1(w) >> for b in t: >> If b =3D=3D 0: >> y =3D SHA1(y) >> If b =3D=3D 1: >> >> y =3D RIPEMD160(y) >> return y >> >> w is a 33-bit stack element, t is a series of 32-bit stack elements >> which we treat as a list of bits. For instance is w=3D312a123e, t=3D0110= 1 >> dGen would return the value created by running >> SHA1(RIPEMD160(SHA1(SHA1(RIPEMD160(SHA1(312a123e)))))) >> >> dGen can be run using OP_SHA1 and OP_RIPEMD160 and can also be run in >> Small Script using Small Script implementations of SHA1 and RIPEMD160. >> Since both sides can use the same stack elements to compute the output >> it lets us build our equivalence check. We just need to find a (w, t) >> such that: >> >> OP_SHA1(A) =3D=3D BigScript-dGen(w, t) // Big script dGen using OP_SHA1 >> and OP_RIPEMD160 >> SHA1(B) =3D=3D SmallScript-dGen(w, t) // Small script dGen >> >> Finding (w,t) is the main computational expense of spending our >> covenant as our covenant depends on this equivalence check. That said >> finding collisions in 160-bit hash functions requires significantly >> less computation than the Bitcoin network regularly performs. See our >> paper for our collision algorithm and discussions of known weaknesses >> in SHA1. >> >> How do you turn this equivalence check into a covenant? >> >> Past work -- https://www.wpsoftware.net/andrew/blog/cat-and-schnorr-tric= ks-i.html >> -- has shown that by structuring a Schnorr signature correctly, that >> Schnorr signature will be the hash of the spending transaction (see >> our paper for how we adapt this to our setting). At very high level >> our covenant works as follows: >> >> 1. Get sighash of spending transaction onto stack >> 2. Use equivalence check to show that small script encoded sighash is >> the same as the sighash we have on the stack >> 3. Open the sighash in small script by showing bytes pushed on stack >> by the spending transaction hash to sighash are bytes of the spending >> transaction. Now we have the bytes of the spending transaction on the >> stack. >> 4. Use Small Script to enforce covenant on the bytes of the spending >> transaction. >> >> See paper for full details. Thanks, >> Ethan > > -- > 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/fc031fe3-444c-446d-a5c1-f2d7a430b478n%40googlegroups.com. --=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/= CAEM%3Dy%2BV2YqRZhvNrgoxMbmaz%3Di94%3DhkAj%3D5HaTErKsOzYq6R4w%40mail.gmail.= com.