| Internet-Draft | The GNU Taler Protocol | August 2025 |
| Gütschow | Expires 13 February 2026 | [Page] |
[ TBW ]¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 13 February 2026.¶
Copyright (c) 2025 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document.¶
[ TBW ]¶
Beware that this document is still work-in-progress and may contain errors. Use at your own risk!¶
"abc" denotes the literal string abc encoded as ASCII [RFC20]¶
a | b denotes the concatenation of a with b¶
len(a) denotes the length in bytes of the byte string a¶
padZero(y, a) denotes the byte string a, zero-padded to the length of y bytes¶
bits(x)/bytes(x) denotes the minimal number of bits/bytes necessary to represent the multiple precision integer x¶
uint(y, x) denotes the y least significant bits of the integer x encoded in network byte order (big endian)¶
uint16(x)/uint32(x)/uint64(x)/uint256(x)/uint512(x) is equivalent to uint(16, x)/uint(32, x)/uint(64, x)/uint(256, x)/uint(512, x), respectively¶
random(y) denotes a randomly generated sequence of y bits¶
a * b (mod N) / a ** b (mod N) denotes the multiplication / exponentiation of multiple precision integers a and b, modulo N¶
SHA-256(msg) -> hash
Input:
msg input message of length L < 2^61 octets
Output:
hash message digest of fixed length HashLen = 32 octets
¶
hash is the output of SHA-256 as per Sections 4.1, 5.1, 6.1, and 6.2 of [RFC6234].¶
SHA-512(msg) -> hash
Input:
msg input message of length L < 2^125 octets
Output:
hash message digest of fixed length HashLen = 64 octets
¶
hash is the output of SHA-512 as per Sections 4.2, 5.2, 6.3, and 6.4 of [RFC6234].¶
SHA-512-256(msg) -> hash
Input:
msg input message of length L < 2^125 octets
Output:
hash message digest of fixed length HashLen = 32 octets
¶
The output hash corresponds to the first 32 octets of the output of SHA-512 defined in Section 3.1.2:¶
temp = SHA-512(msg) hash = temp[0:31]¶
Note that this operation differs from SHA-512/256 as defined in [SHS] in the initial hash value.¶
The Hashed Key Derivation Function (HKDF) used in Taler is an instantiation of [RFC5869]
with two different hash functions for the Extract and Expand step as suggested in [HKDF]:
HKDF-Extract uses HMAC-SHA512, while HKDF-Expand uses HMAC-SHA256 (cf. Section 3.2.1).¶
HKDF(salt, IKM, info, L) -> OKM
Inputs:
salt optional salt value (a non-secret random value);
if not provided, it is set to a string of 64 zeros.
IKM input keying material
info optional context and application specific information
(can be a zero-length string)
L length of output keying material in octets
(<= 255*32 = 8160)
Output:
OKM output keying material (of L octets)
¶
The output OKM is calculated as follows:¶
PRK = HKDF-Extract(salt, IKM) with Hash = SHA-512 (HashLen = 64) OKM = HKDF-Expand(PRK, info, L) with Hash = SHA-256 (HashLen = 32)¶
Based on the HKDF defined in Section 3.3.1, this function returns an OKM that is smaller than a given multiple precision integer N.¶
HKDF-Mod(N, salt, IKM, info) -> OKM
Inputs:
N multiple precision integer
salt optional salt value (a non-secret random value);
if not provided, it is set to a string of 64 zeros.
IKM input keying material
info optional context and application specific information
(can be a zero-length string)
Output:
OKM output keying material (smaller than N)
¶
The final output OKM is determined deterministically based on a counter initialized at zero.¶
counter = 0
do until OKM < N:
x = HKDF(salt, IKM, info | uint16(counter), bytes(N))
OKM = uint(bits(N), x)
counter += 1
¶
RSA-FDH(msg, pubkey) -> fdh
Inputs:
msg message
pubkey RSA public key consisting of modulus N and public exponent e
Output:
fdh full-domain hash of msg over pubkey.N
¶
fdh is calculated based on HKDF-Mod from Section 3.3.2 as follows:¶
info = "RSA-FDA FTpsW!"
salt = uint16(bytes(pubkey.N)) | uint16(bytes(pubkey.e))
| pubkey.N | pubkey.e
fdh = HKDF-Mod(pubkey.N, salt, msg, info)
¶
The resulting fdh can be used to test against a malicious RSA pubkey
by verifying that the greatest common denominator (gcd) of fdh and pubkey.N is 1.¶
RSA-FDH-Derive(bks, pubkey) -> out
Inputs:
bks blinding key secret of length L = 32 octets
pubkey RSA public key consisting of modulus N and public exponent e
Output:
out full-domain hash of bks over pubkey.N
¶
out is calculated based on HKDF-Mod from Section 3.3.2 as follows:¶
info = "Blinding KDF" salt = "Blinding KDF extractor HMAC key" fdh = HKDF-Mod(pubkey.N, salt, bks, info)¶
RSA-FDH-Blind(msg, bks, pubkey) -> out
Inputs:
msg message
bks blinding key secret of length L = 32 octets
pubkey RSA public key consisting of modulus N and public exponent e
Output:
out message blinded for pubkey
¶
out is calculated based on RSA-FDH from Section 3.5.1 as follows:¶
data = RSA-FDH(msg, pubkey) r = RSA-FDH-Derive(bks, pubkey) r_e = r ** pubkey.e (mod pubkey.N) out = r_e * data (mod pubkey.N)¶
RSA-FDH-Sign(data, privkey) -> sig
Inputs:
data data to be signed, an integer smaller than privkey.N
privkey RSA private key consisting of modulus N and private exponent d
Output:
sig signature on data by privkey
¶
sig is calculated as follows:¶
sig = data ** privkey.d (mod privkey.N)¶
RSA-FDH-Unblind(sig, bks, pubkey) -> out
Inputs:
sig blind signature
bks blinding key secret of length L = 32 octets
pubkey RSA public key consisting of modulus N and public exponent e
Output:
out unblinded signature
¶
out is calculated as follows:¶
r = RSA-FDH-Derive(bks, pubkey) r_inv = inverse of r (mod pubkey.N) out = sig * r_inv (mod pubkey.N)¶
RSA-FDH-Verify(msg, sig, pubkey) -> out
Inputs:
msg message
sig signature of pubkey over msg
pubkey RSA public key consisting of modulus N and public exponent e
Output:
out true, if sig is a valid signature
¶
out is calculated based on RSA-FDH from Section 3.5.1 as follows:¶
data = RSA-FDH(msg, pubkey) exp = sig ** pubkey.e (mod pubkey.N) out = (data == exp)¶
Amounts are represented in Taler as positive fixed-point values
consisting of value as the non-negative integer part of the base currency,
the fraction given in units of one hundred millionth (1e-8) of the base currency,
and currency as the 3-11 ASCII characters identifying the currency.¶
Whenever used in the protocol, the binary representation of an amount is
uint64(amount.value) | uint32(amount.fraction) | padZero(12, amount.currency).¶
Absolute timestamps are represented as uint64(x) where x corresponds to
the microseconds since 1970-01-01 00:00 CEST (the UNIX epoch).
The special value 0xFFFFFFFFFFFFFFFF represents "never".¶
All messages to be signed in Taler start with a header containing their size and a fixed signing context (purpose) as registered by GANA in the GNUnet Signature Purposes registry. All Taler-related purposes start at 1000 and can be found in the registry sources.¶
Sign-Msg(purpose, msg) -> out
Inputs:
purpose signature purpose as registered at GANA
msg message content (excl. header) to be signed
Output:
out complete message (incl. header) to be signed
¶
out is formed as follows:¶
out = uint32(len(msg)) | uint32(purpose) | msg¶
// todo: explain persist, check, knows, sum¶
The wallet generates n > 0 coins (coin[i]) and requests n signatures (blind_sig[i]) from the exchange,
attributing value to the coins according to n chosen denominations (denom[i]).
The total value and withdrawal fee (defined by the exchange per denomination)
must be smaller or equal to the amount stored in the single reserve used for withdrawal.¶
wallet exchange
knows denom[i].pub knows denom[i].priv
| |
+----------------------------+ |
| (1) reserve key generation | |
+----------------------------+ |
| |
|----------- (bank transfer) ----------->|
| (subject: reserve.pub, amount: value) |
| |
| +------------------------------+
| | persist (reserve.pub, value) |
| +------------------------------+
| |
+----------------------------------+ |
| (2) coin generation and blinding | |
+----------------------------------+ |
| |
|-------------- /withdraw -------------->|
| (reserve.pub, coin[i].h_denom, |
| blind_coin[i], sig0) |
| |
| +-------------------------------+
| | (3) coin issuance and signing |
| +-------------------------------+
| |
|<----------- blind_sig[i] --------------|
| |
+---------------------+ |
| (4) coin unblinding | |
+---------------------+ |
¶
where (for RSA, without age-restriction)¶
(1) reserve key generation (wallet) reserve = EdDSA-Keygen() persist (reserve, value)¶
The wallet derives coins and blinding secrets using a HKDF from a master secret per withdrawal and an integer index. This is strictly speaking an implementation detail since the master secret is never revealed to any other party, and might be chosen to be implemented differently.¶
(2) coin generation and blinding (wallet)
master_secret = random(256)
persist master_secret
coin_seed[i] = HKDF(uint32(i), master_secret, "taler-withdrawal-coin-derivation", 64)
blind_secret[i] = coin_seed[i][32:]
coin[i].priv = coin_seed[i][:32]
coin[i].pub = EdDSA-GetPub(coin[i].priv)
coin[i].h_denom = SHA-512(uint32(0) | uint32(1) | denom[i].pub)
blind_coin[i] = RSA-FDH-Blind(SHA-512(coin[i].pub), blind_secret[i], denom[i].pub)
msg0 = Sign-Msg(WALLET_RESERVE_WITHDRAW,
( sum(denom[i].value) | sum(denom[i].fee_withdraw)
| SHA-512( SHA-512(denom[i].pub) | uint32(0x1) | blind_coin[i] )
| uint256(0x0) | uint32(0x0) | uint32(0x0) ))
sig0 = EdDSA-Sign(reserve.priv, msg0)
¶
(3) coin issuance and signing (exchange) denom[i] = Denom-Lookup(coin[i].h_denom) check denom[i].pub known and not withdrawal-expired check EdDSA-Verify(reserve.pub, msg, sig) check reserve KYC status ok or not needed check reserve.balance >= sum(denom[i].value + denom[i].fee_withdraw) reserve.balance -= sum(denom[i].value + denom[i].fee_withdraw) blind_sig[i] = RSA-FDH-Sign(blind_coin[i], denom[i].priv) persist withdrawal¶
(4) coin unblinding (wallet) coin[i].sig = RSA-FDH-Unblind(blind_sig[i], blind_secret[i], denom[i].pub) check RSA-FDH-Verify(SHA-512(coin[i].pub), coin[i].sig, denom[i].pub) persist (coin[i], blind_secret[i])¶
wallet merchant
knows merchant.pub knows merchant.priv
knows valid coin[i] knows exchange, payto
| |
| +----------------------+
| | (1) order generation |
| +----------------------+
| |
|<-------- (order.{id,token?}) ----------|
| |
+----------------------+ |
| (2) nonce generation | |
+----------------------+ |
| |
|------- /orders/{order.id}/claim ------>|
| (nonce.pub, token?) |
| |
| +-------------------------+
| | (3) contract generation |
| +-------------------------+
| |
|<----------- contract, sig -------------|
| |
+-------------------------+ |
| (4) payment preparation | |
+-------------------------+ |
| |
|------- /orders/{order.id}/pay -------->|
| (deposit[i]) |
| |
| +-------------------+
| | (5) deposit check |
| +-------------------+
| |
|<---------------- sig ------------------|
| |
+--------------------------+ |
| (6) payment verification | |
+--------------------------+ |
¶
where (without age restriction, policy and wallet data hash)¶
(1) order generation (merchant) wire_salt = random(128) persist order = (id, price, info, token?, wire_salt)¶
(2) nonce generation (wallet) nonce = EdDSA-Keygen() persist nonce.priv¶
Note that the private key of nonce is currently not used anywhere in the protocol.
However, it could be used in the future to prove ownership of an order transaction,
enabling use-cases such as "unclaiming" or transferring an order to another person,
or proving the payment without resorting to the individual coins.¶
(3) contract generation (merchant)
h_wire = HKDF(wire_salt, payto, "merchant-wire-signature", 64)
determine timestamp, refund_deadline, wire_deadline
contract = (order.{id,price,info,token?}, exchange, h_wire, timestamp, refund_deadline, wire_deadline)
check contract.order.token = token?
contract.nonce = nonce.pub
persist contract
h_contract = SHA-512(canonicalJSON(contract))
msg = Sign-Msg(MERCHANT_CONTRACT, h_contract)
sig = EdDSA-Sign(merchant.priv, msg)
¶
(4) payment preparation (wallet)
check EdDSA-Verify(merchant.pub, msg, sig)
check contract.nonce = nonce
TODO: double-check extra hash check?
deposit = CoinSelection(contract.{exchange,price}) TODO: include MarkDirty here
msg[i] = Sign-Msg(WALLET_COIN_DEPOSIT,
( h_contract | uint256(0x0)
| uint512(0x0) | contract.h_wire | coin[i].h_denom
| contract.timestamp | contract.refund_deadline
| deposit[i] + denom[i].fee_deposit
| denom[i].fee_deposit | merchant.pub | uint512(0x0) ))
sig[i] = EdDSA-Sign(coin[i].priv, msg[i])
deposit[i] = (coin[i].{pub,sig,h_denom}, fraction[i], sig[i])
persist (contract, sig[i], deposit[i])
¶
(5) deposit check (merchant) check sum(deposit[i].fraction) == contract.price check Deposit(deposit)[i] msg = Sign-Msg(MERCHANT_PAYMENT_OK, h_contract) sig = EdDSA-Sign(merchant.priv, msg)¶
(6) payment verification check EdDSA-Verify(merchant.pub, msg, sig)¶
merchant/wallet exchange
knows exchange.pub knows exchange.priv
knows merchant.pub knows *denom.pub
knows payto, wire_salt, sig2, h_contract
knows contract, *deposit from Payment
| |
|----------- /batch-deposit ------------>|
| (contract.{timestamp,wire_deadline,refund_deadline}
| h_contract, merchant.pub, sig2, |
| payto, *deposit, wire_salt) |
| |
| *denom = Denom-Lookup(*deposit.coin.h_denom)
| check *denom.pub known and not deposit-expired
| contract.h_wire = HKDF(wire_salt, payto,
| "merchant-wire-signature", 64)
| check *EdDSA-Verify(deposit.coin.pub, msg1, sig1)
| check *RSA-FDH-Verify(SHA-512(deposit.coin.pub),
| deposit.coin.sig, denom.pub)
| check not overspending
| persist deposit-record, mark-spent
| schedule bank transfer
| exchange_timestamp = now()
| sig3 = EdDSA-Sign(exchange.priv, msg3)
| |
|<----- sig3, exchange_timestamp --------|
| |
check EdDSA-Verify(exchange.pub, msg3, sig3) |
¶
where msg3 is formed as follows:¶
msg3 = Sign-Msg(EXCHANGE_CONFIRM_DEPOSIT,
( h_contract | contract.h_wire | uint512(0x0)
| exchange_timestamp | contract.wire_deadline
| contract.refund_deadline
| sum(*deposit.fraction - *denom.fee_deposit)
| SHA-512(*deposit.sig1) | merchant.pub ))
¶
[ TBD ]¶
None.¶
[ TBD ]¶
This work was supported in part by the German Federal Ministry of Education and Research (BMBF) within the project Concrete Contracts.¶