Adaptor signatures

Also covering Signature adaptors and Scriptless scripts

Adaptor signatures (also called signature adaptors) are auxiliary signature data that commit to a hidden value. When an adaptor is combined with a corresponding signature, it reveals the hidden value. Alternatively, when combined with the hidden value, the adaptor reveals the signature. Other people may create secondary adaptors that reuse the commitment even if they don’t know the hidden value. This makes adaptors a powerful tool for implementing locking in bitcoin contracts.

Contracts in Bitcoin often require a locking mechanism to ensure the atomicity of a set of payments—either all the payments succeed or all of them fail. This locking has traditionally been done by having all payments in the set commit to the same hash digest preimage; when the party who knows the preimage reveals it onchain, everyone else learns it and can unlock their own payments. Commonly used hashlocks in Bitcoin consume about 67 bytes and reveal the link between the set of payments because they all use the same preimage and digest.

By comparison, signature adaptors never need to be published onchain. To anyone without a corresponding adaptor, a signature created with an adaptor looks like any other digital signature, giving adaptors significant efficiency and privacy advantages over hashlocks.

Example

The multiple uses of signature adaptors can be seen in a simple coinswap protocol. For example, Alice can give Bob an adaptor for an unsigned transaction that promises to pay him 1 BTC. An adaptor by itself can’t be used as a BIP340 signature, so Alice hasn’t paid Bob yet.

What the adaptor does provide Bob is a commitment to Alice’s hidden value. This commitment includes a parameter Bob can use to create a second adaptor that commits to the same hidden value as Alice’s adaptor. Bob can make that commitment even without knowing Alice’s hidden value or his own signature for that commitment. Bob gives Alice his adaptor and a corresponding unsigned transaction that promises to pay her 1 BTC.

Alice has always known the hidden value, so she can combine the hidden value with Bob’s adaptor to get his signature for the transaction that pays her. She broadcasts the transaction and receives Bob’s payment. When Bob sees that transaction onchain, he can combine its signature with the adaptor he gave Alice, allowing him to derive the hidden value. Then he can combine that hidden value with the adaptor Alice gave him earlier. Bob broadcasts that transaction to receive Alice’s payment, completing the coinswap.

Besides coinswaps, there are several other proposed uses for adaptor signatures.

Click to display the same coinswap example in mathematical terms

In the following example, we assume the use of BIP340 schnorr signatures. We use lowercase variables for scalars and uppercase variables for elliptic curve points. We represent concatenation with || and the hash function with H().

Alice creates a valid signature commitment (s) for the transaction paying Bob (m) using her private key (p), which corresponds to her public key (P = pG). She also uses a private random nonce (r), a hidden value (t), and the elliptic curve points for them (R = rG, T = tG):

s = r + t + H(R + T || P || m) * p

She subtracts t from the signature commitment to produce a signature adaptor:

s' = s - t

She gives Bob the adaptor, which consists of the following data:

s', R, T

Bob can verify the adaptor:

s' * G ?= R + H(R + T || P || m) * P

But the adaptor is not a valid BIP340 signature. For a valid signature, BIP340 expects x and Y, using them with the expression:

x * G ?= Y + H(Y || P || m) * P

However,

If Bob sets Y = R so that it matches the s' he received in the adaptor, then BIP340 is going to fail on H(R || P || m) since Alice computed her hash with H(R + T || P || m).
If Bob sets Y = R + T so that it matches H(R + T || P || m), BIP340 is going to fail on the initial Y since Bob is providing R + T rather than the needed R.

Therefore Bob can’t use the adaptor as a BIP340 signature. However, he can create his own adaptor using it. This is similar to the signature Alice created but Bob doesn’t commit to t here, since Bob doesn’t know that value. All variables here except T are different for Bob than they were for Alice:

s = r + H(R + T || P || m) * p

Unlike Alice, Bob doesn’t need to tweak his signature. Bob’s signature commitment s is not a part of a valid signature because it commits to r and R + T, which won’t pass BIP340 verification for the same reasons previously described. To be valid, the signature needs to commit to r + t and R + T, which Bob can’t produce since he doesn’t know t.

Bob gives Alice his adaptor:

s, R, T

Alice already knew T, but (s, R, T) is a standard signature adaptor so we use its full form. Alice can produce a signature from that adaptor using the hidden t value that only she knows so far:

(s + t) * G ?= R + T + H(R + T || P || m) * P

Alice uses the signature to broadcast Bob’s transaction that pays her. When Bob sees (s + t) onchain, he can learn the value of t:

t = (s + t) - s

He can then use t to solve the adaptor Alice gave him earlier:

(s' + t) * G ?= R + T + H(R + T || P || m) * P

Bob uses that signature to broadcast the transaction Alice originally gave him.

Relationship to multiparty signatures

Signature adaptors usually can’t secure a contract entirely by themselves. For example, in the above description of a coinswap, Alice could double spend her payment to Bob after she learned Bob’s signature, or Bob could’ve tried the same in reverse (with more difficulty since we assumed Alice’s transaction had one confirmation). This issue is typically addressed by combining signature adaptors with multiparty signatures. For example, Alice deposits her money into an address that can only be spent if both she and Bob collaborate to create a valid signature. Now Alice can provide Bob with an adaptor for her half of the multiparty signature, which Bob can accept with perfect safety knowing that Alice couldn’t double spend the funds without his participation. This may additionally require a timelocked refund option in case one party refuses to sign.

In the schnorr signature scheme, signature adaptors are usually proposed to be combined with multiparty signature schemes such as MuSig to allow the published signature to look like a single-party signature, enhancing privacy and efficiency. This is also possible in ECDSA but it requires novel algorithms that are either comparatively slow or require additional security assumptions. Instead, an alternative scheme for adaptor signatures exists for Bitcoin that uses 2-of-2 OP_CHECKSIG multisig; this is less efficient and possibly less private—but arguably simpler and safer than multiparty ECDSA.

Primary code and documentation

2021

2020

2019

2018

Example

Relationship to multiparty signatures

Primary code and documentation

Optech newsletter and website mentions

See also