This repository is a Typescript interface to Dock's Rust crypto library. It uses the WASM wrapper.
To use this package within your project simply run
npm install @docknetwork/crypto-wasm-ts
Or with Yarn
yarn add @docknetwork/crypto-wasm-ts
To build the project run:
yarn build
To run the all tests in the project run:
yarn test
Following is a conceptual explanation of the primitives.
Disclaimer: There's multiple variations of the BBS scheme available out there, of which BBS+ was thought to be the version that is proven to be secure. However, in a recent paper, it was proven that in the initial BBS design was secure, with even a smaller signature size. This version is also the one used in the IRTF standardization effort.
The BBS signature Scheme (henceforth referred to as just BBS) allows for signing an ordered list of messages, producing a signature of constant size independent of the number
of messages. The signer needs to have a public-private keypair and signature parameters which are public values whose size
depends on the number of messages being signed. A verifier who needs to verify the signature needs to know the
signature parameters used to sign the messages and the public key of the signer. In the context of anonymous credentials,
messages are called attributes.
BBS also allows a user to request a blind signature from a signer where the signer does not know 1 or more messages
from the list. The user can then unblind the blind signature to get a regular signature which can be verified by a verifier in
the usual way. Such blind signatures can be used to hide a user specific secret like a private key or some unique identifier
as a message in the message list and the signer does not become aware of the hidden message.
With a BBS signature, a user in possession of the signature and messages and create a zero-knowledge proof of knowledge
of the signature and the corresponding signed messages such that he can prove to a verifier that he knows a signature and the
messages and optionally reveal one or more of the messages.
A typical use of BBS signatures looks like:
An accumulator is a set-like data structure in which elements can be added or removed but the size of the accumulator remains constant. However, an accumulator cannot be directly checked for presence of an element, an element needs to have accompanying data called the witness (its the manager's signature on the element), the element and the witness and these together can be used to check the presence or absence of the element. An accumulator can be considered similar to the root of the merkle tree where the inclusion proof is the witness of the element (non-membership proofs aren't possible with simple merkle trees). As with merkle trees, as elements are added or removed from the accumulator, the witness (inclusion proof) needs to be updated for the current accumulated value (root).
2 kinds of accumulators are provided, positive and universal.
Positive support only membership witnesses while universal support both membership and non-membership witnesses.
Creating non-membership witnesses is expensive however, and the cost depends on the number of members present in the accumulator.
Both accumulators are owned by an accumulator manager who has the private key to the accumulator
and only the owner can add or remove elements or create witnesses using the accumulator.
Accumulator allows proving membership of the member (or non-member) and the corresponding witness in zero knowledge meaning
a user in possession of an accumulator member (or non-member) and the witness can convince a verifier that he knows of an
element present (or absent) in the accumulator without revealing the element or the witness. Note, the like merkle trees,
witnesses (inclusion proof) are tied to the accumulated value (root) and need to be updated as accumulator changes.
Witnesses can be updated either by the accumulator manager using his private key or the manager can publish witness update
information and the updates (additions and removals) and users can update their witnesses.
A typical use of accumulator looks like:
The above primitives can be combined using the composite proof system. An example is (in zero knowledge) proving knowledge of 2 different signatures and the message lists. Another example is proving knowledge of the signature and messages and certain message's presence (absence) in an accumulator. Or the knowledge of 5 signatures and proving certain message is the same in the 5 message lists.
Before calling any function that calls the underlying WASM, use initializeWasm to load the WASM module.
This function returns a promise which is resolved once the WASM module is successfully loaded.
import { initializeWasm } from '@docknetwork/crypto-wasm-ts'
// Load the WASM module
await initializeWasm();
The library has support for BBS, BBS+, PS, and BBDT16 schemes.
Although they're similar in many aspects, specially for BBS and BBS, you're advised to consult the tests for changes
between the different schemes.
However, you can make a good guess just by looking at the schema here.
For all the following examples, BBS will be used, but the concepts should be transferable to other schemes.
BBS signatures sign an ordered list of messages and thus it is important to serialize your signing payload in this format. Eg, in case of a credential with attributes in JSON format where each attribute is a key, convert the JSON to a list of attributes and this conversion should be deterministic, meaning attributes should always end up in the same order. Following is a conversion of a JSON credential with 4 attributes to a list where the values are placed in the alphabetical order of the keys:
Given JSON
{
"ssn": "12345678",
"fname": "John",
"lname": "Smith",
"city": "NYC"
}
Converted to list
["NYC", "John", "Smith", "12345678"]
Now each element of the above list must be converted to bytearrays, i.e. Uint8Array and the type of above list becomes Uint8Array[].
Before messages can be signed, 2 things are needed:
Signature parameters: Public values, that can be created by anyone but must be known to the signer and verifier to sign and verify respectively. To create them, the number of messages (attributes) being signed must be known and the size of the parameters increases with the number. In the above example, number of attributes is 4. These parameters can be generated randomly or deterministically by using a publicly known label. It is advised to use the latter as it allows for extending/shrinking the same parameters when number of messages change.
Keypair: To create and verify BBS signatures, the signer (issuer in case of a credential) needs to create a secret key to sign, public key to verify.
2 ways of generating signature parameters
import { BBSSignatureParams } from '@docknetwork/crypto-wasm-ts';
const messageCount = 4;
// Randomly generated params
const paramsRandom = BBSSignatureParams.generate(messageCount);
// the following function will be useful throughout the documentation
function stringToBytes(message: string): Uint8Array {
return Uint8Array.from(Buffer.from(message, 'utf-8'));
}
const label = stringToBytes("My sig params");
// Deterministically generated params
const paramsDeterministc = BBSSignatureParams.generate(messageCount, label);
// Deterministic params can be extended if messageCount changes, say to 5 or 3
const paramsDeterministc5 = paramsDeterministc.adapt(5);
const paramsDeterministc3 = paramsDeterministc.adapt(3);
Generating a keypair once signature parameters are created.
import { BBSKeypair } from '@docknetwork/crypto-wasm-ts';
const keypair = BBSKeypair.generate(paramsDeterministc);
const sk = keypair.secretKey;
const pk = keypair.publicKey;
Each one of the messages should be a Uint8Array
const messages: Uint8Array[] = ["NYC", "John", "Smith", "12345678"].map(element => stringToBytes(element));
When the messages are arbitrary bytes, they need to be encoded to a field element (a number in certain range). You can either let the signing function encode it by passing
the encode argument as true to encode it using your own encoding function.
Letting the signing function encode
import { BBSSignature } from '@docknetwork/crypto-wasm-ts';
// The signing function will encode bytes to a field element as true is passed
const sig = BBSSignature.generate(messages, sk, paramsDeterministc, true);
// As the messages are not encoded, pass true to the verification function to make it encode messages before verifying the signature.
const result = sig.verify(messages, pk, paramsDeterministc, true);
console.assert(result.verified);
Passing pre-encoded messages to signing function
const encodedMessages = [];
for (let i = 0; i < messages.length; i++) {
encodedMessages.push(generateFieldElementFromBytes(messages[i]));
}
// The signing function will not encode as false is passed
const sig = BBSSignature.generate(encodedMessages, sk, params, false);
// As the messages are pre-encoded, pass false to the verification function to avoid encoding messages before verifying the signature.
const result = sig.verify(encodedMessages, pk, params, false);
console.assert(result.verified);
Proving and verifying knowledge of signature can be done with or without using the composite proof system but this doc will only describe using the composite proof system. For the other way, see tests here
The code for BBS lives here.
Similar to BBS signatures, accumulators also have a setup phase where public parameters and keys are generated and these public values need to be published. The accumulator manager's signing key is needed to update the accumulator or create a witness and the public key is needed to verify the (non)membership. This document talks only about Positive accumulator, for universal accumulator see the corresponding tests.
Similar to BBS, parameters can be generated randomly or deterministically.
// Randomly generated params
const paramsRandom = PositiveAccumulator.generateParams();
const label = stringToBytes("My sig params");
// Deterministically generated params
const params = PositiveAccumulator.generateParams(label);
Generating a keypair once parameters are created.
const keypair = PositiveAccumulator.generateKeypair(params);
Initialize the accumulator
const accumulator = PositiveAccumulator.initialize(params);
Care must be taken to not add duplicate elements in the accumulator or remove non-existent elements or creating witness of
non-existing elements. The accumulator itself cannot make such checks and thus this state must be tracked separately.
The interface for such a state is IAccumulatorState. Its strongly recommended that
this state should be passed as an argument to the add, remove, and other functions that expect it. However, it's not mandatory
as the caller might have its own way of avoiding such issues. The tests below use an in-memory state InMemoryState which
implements IAccumulatorState interface.
Elements can be added/removed one by one or in a batch. Before adding an element, it must be encoded to a field element.
Encoding a positive integer can be done using encodePositiveNumberAsAccumulatorMember, arbitrary bytes can be encoded as
encodeBytesAsAccumulatorMember.
Adding 2 elements in the accumulator
const state = new InMemoryState();
const e1 = Accumulator.encodePositiveNumberAsAccumulatorMember(101);
const bytes: Uint8Array = [...];
const e2 = Accumulator.encodeBytesAsAccumulatorMember(bytes);
await accumulator.add(e1, sk, state);
await accumulator.add(e2, sk, state);
Removing an existing element
await accumulator.remove(e2, sk, state);
Adding multiple elements in a batch
const e3 = Accumulator.encodePositiveNumberAsAccumulatorMember(103);
const e4 = Accumulator.encodePositiveNumberAsAccumulatorMember(104);
await accumulator.addBatch([e3, e4], sk, state);
Adding and removing multiple elements in a batch
// Elements to add
const additions: Uint8Array[] = [...];
// Elements to remove
const removals: Uint8Array[] = [...];
await accumulator.addRemoveBatches(additions, removals, sk, state);
Once an element is added to the accumulator by the manager, a witness is required to verify the membership. Also required is the accumulator value when the witness was created, this value should be publicly available.
Generating a membership witness
// Note that the secret key is needed to create the witness
const witness = await accumulator.membershipWitness(e4, sk, state)
Verify the membership
// The accumulated value `accumulator.accumulated` is posted publicly
const verifAccumulator = PositiveAccumulator.fromAccumulated(accumulator.accumulated);
// Note that only public values needed to verify the membership
console.assert(verifAccumulator.verifyMembershipWitness(e4, witness, pk, params));
As the accumulator changes, the witness needs to be updated as well. The witness can be updated without the manager's help if the updates (additions, removals) are known.
Update witness after an addition
// Say element e8 was added after the witness of e4 was created
witness.updatePostAdd(e8, e4, accumulator.accumulated);
Update witness after a removal
ts // Say element e1 was removed after the witness of e4 was created witness.updatePostRemove(e1, e4, accumulator.accumulated);
The above method of updating the witness by going over each update is slow. The manager can make this process more efficient
for all members by publishing a WitnessUpdatePublicInfo built using the updates to the older accumulator. All members can then
use this public information to update their witnesses.
Manager creates WitnessUpdatePublicInfo and then updates the accumulator
// The current accumulated value is accumulator.accumulated
// Elements to add
const additions: Uint8Array[] = [...];
// Elements to remove
const removals: Uint8Array[] = [...];
// This will be published along with `additions` and `removals`
const witnessUpdInfo = WitnessUpdatePublicInfo.new(accumulator.accumulated, additions, removals, sk);
// Update the accumulator now
await accumulator.addRemoveBatches(additions, removals, sk, state);
The member can now fetch the update information and update as
witness.updateUsingPublicInfoPostBatchUpdate(e4, additions, removals, witnessUpdInfo);
The member can update his witness given multiple such updates using updateUsingPublicInfoPostMultipleBatchUpdates. See the tests for examples.
The above workflow requires that after every addition to the accumulator, the new accumulator must be published along with witness update info so that other members can update their witnesses. This is however expensive as the accumulator and update info might be posted on the blockchain and also every existing member has to update its witness. One way to mitigate that is to create pre-filled accumulators meaning that before publishing the accumulator the first time, the manager adds all the member ids in the accumulator. This strategy assumes that member ids are either predictable like monotonically increasing numbers or the manager can internally keep a map of random ids like UUIDs to a number. Now when the manager actually wants to allow a member to prove membership, he can create a witness for that member but the accumulator value remains same and thus the witness for existing members also remain same. It should be noted though that changing the accumulator value causes change in all existing witnesses and thus its better to make a good estimate of the number of members during prefill stage. See this test for a complete example using a positive accumulator.
Proof of membership and non-memberships can be done with or without using the composite proof system but this doc will only describe using the composite proof system.
The code for accumulators lives here.
Statement.Statement index and 2nd item is index of that data point in the witness.MetaStatement will refer to the WitnessRef of each attribute. This is
public information as well.StatementsStatements, MetaStatements and an optional context. Both prover and verifier should independently create this.
The prover uses the ProofSpec and all Witnesses to create the proof and the verifier uses theProofSpec to verify the proof.A complete example is shown in this test.
Proving knowledge of 1 BBS signature over the attributes and only disclosing some attributes. Say there are 5 attributes in the credential: SSN, first name, last name, email and city, and they are present in the attribute list in that order. The prover wants to reveal his last name and city, but not any other attribute while proving that he possesses such a credential signed by the issuer.
// The attributes, [SSN, first name, last name, email, city]
const messages: Uint8Array[] = [...];
// Public values
const params: BBSSignatureParams;
const pk: BBSPublicKey;
// The signature
const sig: BBSSignature = ...;
// Prover prepares the attributes he wants to disclose, i.e. attribute index 2 and 4 (indexing is 0-based), and the ones he wants to hide.
const revealedMsgIndices: Set<number> = new Set();
revealedMsgIndices.add(2);
revealedMsgIndices.add(4);
// revealedMsgs are the attributes disclosed to the verifier
const revealedMsgs: Map<number, Uint8Array> = new Map();
revealedMsgs.set(2, messages[2]);
// unrevealedMsgs are the attributes hidden from the verifier
const unrevealedMsgs: Map<number, Uint8Array> = new Map();
unrevealedMsgs.set(0, messages[0]);
unrevealedMsgs.set(1, messages[1]);
unrevealedMsgs.set(3, messages[3]);
Since there is only 1 kind of proof, i.e. the knowledge of a BBS signature and the signed attributes, there would be only 1 Statement.
import { Statement, Statements } from '@docknetwork/crypto-wasm-ts'
// Create a BBS signature, true indicates that attributes/messages are arbitrary bytes and should be encoded first
const statement1 = Statement.bbsSignatureProverConstantTime(paramsDeterministc, revealedMsgs, true);
const statements = new Statements();
statements.add(statement1);
// Optional context of the proof, this can specify the reason why the proof was created or date of the proof, or self-attested attributes (as JSON string), etc
const context = stringToBytes('some context');
Once it has been established what needs to be proven, ProofSpec needs to be created which represents all the requirements.
Both the prover and verifier should independently construct this ProofSpec. Note that there are no MetaStatements as there are no
other conditions on the witnesses and thus its empty
import { ProofSpec, MetaStatements } from '@docknetwork/crypto-wasm-ts';
const ms = new MetaStatements();
const proofSpec = new ProofSpec(statements, ms, [], context);
Prover creates Witness using the signature and hidden attributes
import { Witness, Witnesses } from '@docknetwork/crypto-wasm-ts';
const witness1 = Witness.bbsSignatureConstantTime(sig, unrevealedMsgs, true);
const witnesses = new Witnesses();
witnesses.add(witness1);
Prover now uses the ProofSpec to create the proof. To ensure that the prover is not replaying, i.e. reusing a proof created by someone else, the verifier can request the prover to include its provided nonce in the proof.
import { CompositeProof } from '@docknetwork/crypto-wasm-ts';
const nonce = stringToBytes('a unique nonce given by verifier');
const proof = CompositeProof.generate(proofSpec, witnesses, nonce);
Verifier can now verify this proof. Note that the verifier does not and must not receive ProofSpec from prover, it
needs to generate on its own.
console.assert(proof.verify(proofSpec, nonce).verified);
The examples shown here have assumed that the number of messages for given signature params is fixed but that might not be always true. An example is where some of the messages in the signature are null (like N/A) in certain signatures. Eg, when the messages are attributes in a credential that specifies the educational qualifications and institutes of a person, someone with a high school level education will have N/A for attributes like university name, major, etc. One way to deal with it is to decide some sentinel value like 0 for all the N/A attributes and disclose those attributes while creating a proof. Other is to have certain attribute in the credential specify which attribute indices that are N/A and always reveal this attribute. A complete example of the latter is shown in this test.
A complete example is shown in this test.
Proving knowledge of 2 BBS signatures over the attributes and only disclosing some attribute and proving equality of 1 attribute without disclosing it. Say there are 2 credentials and hence 2 BBS signatures. One credential has 5 attributes: SSN, first name, last name, email and city and the other has 6 attributes name, email, city, employer, employee id and SSN and in that order. The prover wants to prove that he has those 2 credentials, reveal his employer name and prove that SSN in both credentials is the same without revealing the SSN.
// The attributes from 1st credential, [SSN, first name, last name, email, city]
const messages1: Uint8Array[] = [...];
// The attributes from 2nd credential, [name, email, city, employer, employee id, SSN]
const messages2: Uint8Array[] = [...];
// Public values for 1st issuer
const parasm1: BBSSignatureParams;
const pk1: BBSPublicKey;
// Public values for 2nd issuer
const parasm2: BBSSignatureParams;
const pk2: BBSPublicKey;
// The signature from 1st credential
const sig1: BBSSignature = ...;
// The signature from 2nd credential
const sig2: BBSSignature = ...;
Since the prover is proving possession of 2 BBS signatures, there will be 2 Statements. Also, for the 2nd signature prover is
revealing employer attribute, which is at index 3.
// Statement for signature of 1st signer, not revealing any messages to the verifier
const statement1 = Statement.bbsSignatureProverConstantTime(params1, new Map(), true);
// Statement for signature of 2nd signer, revealing 1 message to the verifier
const revealedMsgIndices: Set<number> = new Set();
revealedMsgIndices.add(3);
const revealedMsgs: Map<number, Uint8Array> = new Map();
const unrevealedMsgs2: Map<number, Uint8Array> = new Map();
for (let i = 0; i < messageCount2; i++) {
if (revealedMsgIndices.has(i)) {
revealedMsgs.set(i, messages2[i]);
} else {
unrevealedMsgs2.set(i, messages2[i]);
}
}
const statement2 = Statement.bbsSignatureProverConstantTime(params2, revealedMsgs, true);
// Collect all the statements
const statements = new Statements();
const sId1 = statements.add(statement1);
const sId2 = statements.add(statement2);
The prover has 2 prove that both credentials contain the same SSN which is same as saying for the 1st signature (1st Statement),
attribute at index 0 is equal to 2nd signature's (2nd Statement) attribute index 5. This requires the use of a MetaStatement to
express this condition, specifically MetaStatement.witnessEquality which takes the WitnessRef for each witness that needs to be
proven equal. WitnessRef for SSN in 1st signature is (0, 0) and in 2nd signature is (1, 5). Create a WitnessEqualityMetaStatement to express that.
// For proving equality of SSN, messages1[0] == messages2[5], specify using MetaStatement
const witnessEq = new WitnessEqualityMetaStatement();
witnessEq.addWitnessRef(0, 0);
witnessEq.addWitnessRef(1, 5);
const ms = MetaStatement.witnessEquality(witnessEq);
const metaStatements = new MetaStatements();
metaStatements.add(ms);
Incase equality of additional attribute also needs to be proven say email, then WitnessEqualityMetaStatement needs to be created
for the WitnessRef of email in both signatures.
// For proving equality of email, messages1[3] == messages2[1], specify using MetaStatement
const witnessEq2 = new WitnessEqualityMetaStatement();
witnessEq2.addWitnessRef(sId1, 3);
witnessEq2.addWitnessRef(sId2, 1);
const ms2 = MetaStatement.witnessEquality(witnessEq2);
metaStatements.add(ms2);
Similar to before, once it has been established what needs to be proven, ProofSpec needs to be created with all Statementss and MetaStatements.
const proofSpec = new ProofSpec(statements, metaStatements);
The prover creates the witnesses with both signatures and messages that he is hiding from the verifier
// Using the messages and signature from 1st signer
const unrevealedMsgs1 = new Map(messages1.map((m, i) => [i, m]));
const witness1 = Witness.bbsSignatureConstantTime(sig1, unrevealedMsgs1, true);
// Using the messages and signature from 2nd signer
const witness2 = Witness.bbsSignatureConstantTime(sig2, unrevealedMsgs2, true);
const witnesses = new Witnesses();
witnesses.add(witness1);
witnesses.add(witness2);
const proof = CompositeProof.generate(proofSpec, witnesses);
Verifier verifies the proof.
console.assert(proof.verify(proofSpec).verified);
Say a prover has a credential where one of the attribute is added to an accumulator. The prover wants to prove that his attribute is a member of the accumulator without revealing the attribute itself. Say the attributes are SSN, first name, last name, email and user-id and the prover wants to prove that the user-id is present in the accumulator without revealing it to the verifier.
// The attributes, [SSN, first name, last name, email, user-id]
const messages: Uint8Array[] = [...];
Because the attributes for accumulator and BBS signatures are encoded differently, attributes are pre-encoded.
// Encode messages for signing as well as adding to the accumulator
const encodedMessages = [];
for (let i = 0; i < messageCount; i++) {
if (i === messageCount-1) {
// Last one, i.e. user id is added to the accumulator so encode accordingly
encodedMessages.push(Accumulator.encodeBytesAsAccumulatorMember(messages[i]));
} else {
encodedMessages.push(Signature.encodeMessageForSigning(messages[i]));
}
}
Both signer and accumulator manager will have public params and their secret keys
const sigParams = BBSSignatureParams.generate(5, label);
// Signers keys
const sigSk: BBSPlusSecretKey = ...;
const sigPk: BBSPublicKey = ...;
// Accumulator manager's params, keys and state
const accumParams = PositiveAccumulator.generateParams(stringToBytes('Accumulator params'));
const accumKeypair = PositiveAccumulator.generateKeypair(accumParams);
const accumulator = PositiveAccumulator.initialize(accumParams);
const state = new InMemoryState();
Signer signs the credential and accumulator manager adds the attribute to the credential and sends the witness to the prover
// Signer signs the message
const sig: BBSSignature = ...;
// user-id is at index 4 is the message list
const userIdIdx = 4;
await accumulator.add(encodedMessages[userIdIdx], accumKeypair.secret_key, state);
const accumWitness = await accumulator.membershipWitness(encodedMessages[userIdIdx], accumKeypair.secret_key, state)
To prove accumulator membership in zero-knowledge, the prover and verifier agree on set of public parameters called the ProvingKey.
This is not specific to the accumulator and can be reused for any number of accumulators. Also a prover might use different
proving keys when interacting with different verifiers. Its recommended generating the proving key deterministically by passing a label.
const provingKey = Accumulator.generateMembershipProvingKey(stringToBytes('Our proving key'));
The prover needs to prove 2 Statements, knowledge of BBS signature and knowledge of accumulator member and corresponding witness.
const statement1 = Statement.bbsSignatureProverConstantTime(sigParams, revealedMsgs, false);
const statement2 = Statement.accumulatorMembership(accumParams, accumKeypair.public_key, provingKey, accumulator.accumulated);
const statements = new Statements();
statements.add(statement1);
statements.add(statement2);
The prover also needs to prove that the accumulator member is same as the credential attribute at index 4, the user id.
The WitnessRef of the accumulator member is (1, 0) as index of membership Statement is 1 and index of member is always 0.
// The last message in the signature is same as the accumulator member
const witnessEq = new WitnessEqualityMetaStatement();
// Witness ref for last message in the signature
witnessEq.addWitnessRef(0, userIdIdx);
// Witness ref for accumulator member
witnessEq.addWitnessRef(1, 0);
const ms = MetaStatement.witnessEquality(witnessEq);
const metaStatements = new MetaStatements();
metaStatements.add(ms);
const proofSpec = new ProofSpec(statements, metaStatements);
The prover creates Witnesses for all statements and then creates the proof. The Witness for Statement.accumulatorMembership contains
the member and the accumulator witness.
const witness1 = Witness.bbsSignatureConstantTime(sig, unrevealedMsgs, false);
const witness2 = Witness.accumulatorMembership(encodedMessages[userIdIdx], accumWitness);
const witnesses = new Witnesses();
witnesses.add(witness1);
witnesses.add(witness2);
const proof = CompositeProof.generate(proofSpec, witnesses);
Disclaimer: With BBS, there is no blinding (the commitment is computationally hiding, though it can be made perfectly hiding by adding a dummy attribute). However, it can be easily achieved with BBS+. A complete example is shown in this test.
A signature is blind when the signer is not aware of the message (or a part of the message) that he is signing, thus the signer is blind. Blind signature in credential is used when the holder does not want the signer to learn some attribute, eg. one of the credential attribute is a secret key and the holder does not want the signer to learn the secret key. Here the user creates a commitment to the "blinded", i.e. hidden attributes to convince the signer that he is only hiding the certain attribute(s). Eg if a credential has 5 attributes secret1, name, secret2, email, city and the user wants to hide secret1 and secret2 from the signer, the signer wants to be sure that the user is indeed hiding attributes at index 0 and 2, not others. The prover uses the composite proof system to prove that he knows that the commitment contains those 2 attributes
// Messages are secret1, name, secret2, email, city
// Signature params for 5 attributes
const sigParams = BBSPlusSignatureParamsG1.generate(5, label);
// Prepare messages that will be blinded (hidden) and known to signer
const blindedMessages = new Map();
// User wants to hide messages at indices 0 and 2 from signer
const blindedIndices: number[] = [];
blindedIndices.push(0);
blindedMessages.set(0, stringToBytes('my-secret'));
blindedIndices.push(2);
blindedMessages.set(2, stringToBytes('my-another-secret'));
The signature requester, prover in this case, creates a blind signature request. In addition to the request, it also returns
randomness blinding that goes into the commitment. This randomness is later used
import { BBSPlusBlindSignatureG1 } from '@docknetwork/crypto-wasm-ts';
// Blind signature request will contain a commitment,
const [blinding, request] = BBSPlusBlindSignatureG1.generateRequest(blindedMessages, params, true);
The proof needs to be over only 1 Statement, the statement proving knowledge of the committed attributes in the commitment.
To create the commitment, a commitment key (public values) needs to be created from the signature params
// Take parts of the sig params corresponding to the blinded messages and create the commitment key commKey
const commKey = params.getParamsForIndices(request.blindedIndices);
const statement1 = Statement.pedersenCommitmentG1(bases, request.commitment);
const statements = new Statements();
statements.add(statement1);
const proofSpec = new ProofSpec(statements, new MetaStatements());
Now the prover creates witness for the commitment Statement using the randomness and the hidden attributes.
import { getBBSWitnessForBlindSigRequest } from '@docknetwork/crypto-wasm-ts';
const witness1 = getBBSWitnessForBlindSigRequest(blindedMessages)
const witnesses = new Witnesses();
witnesses.add(witness1);
const proof = CompositeProof.generate(proofSpec, witnesses);
Signer now verifies the proof. Note that the signer independently creates the ProofSpec as he knows which attributes are being
hidden from him. If the proof is correct, signer creates a blind signature using the known attributes and the commitment
and sends to the prover.
console.assert(proof.verify(proofSpec).verified);
// Signer is convinced that user knows the opening to the commitment
// Signer creates a blind signature with these revealed messages and the commitment.
revealedMessages.set(1, stringToBytes('John Smith'));
revealedMessages.set(3, stringToBytes('john.smith@emample.com'));
revealedMessages.set(4, stringToBytes('New York'));
const blindSig = BBSPlusBlindSignatureG1.generate(request.commitment, revealedMessages, sk, params, true);
The prover can now "unblind" the signature meaning he can convert a blind signature into a regular BBS signature which he can use in proof as shown in examples above
// Unbling the signature from the randomness of the commitment.
const sig = blindSig.unblind(blinding);
// Combine blinded and revealed messages in an array
const messages = Array(blindedMessages.size + revealedMessages.size);
for (const [i, m] of blindedMessages.entries()) {
messages[i] = m;
}
for (const [i, m] of revealedMessages.entries()) {
messages[i] = m;
}
// Signature can be verified
const result = sig.verify(messages, pk, params, true);
console.assert(result.verified);
A pseudonym is meant to be used as a unique identifier. It can be considered as a public key where the creator of the pseudonym has the secret key, and it can prove the knowledge of this secret key. A pseudonym can also be bound to multiple attributes from multiple credentials. This concept was introduced in Attribute-based Credentials for Trust.
Motivation: Proving knowledge of BBS signatures is unlinkable meaning the verifier cannot link to 2 proofs presented from the same
credential (signature). But this might not always be desirable for the verifier and the prover might agree to being linked for any
proofs that he creates for that particular verifier without revealing any attribute of the credential.
A verifier wants to attach a unique identifier to a prover without either learning anything unintended (by prover)
from the prover's signature nor can that unique identifier be used by other verifiers to identify the prover,
eg. a seller (as a verifier) should be able to identify repeat customers (prover) by using a unique identifier, but
he should not be able to share that unique identifier with other sellers using their own identifier for that prover.
Above is achieved by making the prover go through a one-time registration process with the verifier where the prover creates a pseudonym and shares the pseudonym with the verifier. The prover on subsequent interactions share the pseudonym and proof of knowledge of the pseudonym's secret key with the verifier. Thus, pseudonyms allow for verifier-local and opt-in linkability.
In the test, the credential has 4 attributes, SSN, first name, last name and email and during registration, the prover creates many pseudonyms, for different verifiers, some are bound to attributes and some not. See the test for more details.
A social KYC (Know Your Customer) credential claims that the subject owns certain social media profile like a twitter profile
credential claims that a user owns the twitter profile with certain handle. Here the issuer of the credential must verify
the user's control of the profile. One way to achieve that is for the user to post a unique issuer supplied challenge string
on his profile, like tweeting it when requesting twitter profile credential. This makes the process 2-step, in step 1 user
requests the challenge from issuer and which he tweets and in step 2, he asks the issuer to check the tweet and issue him
a credential. An alternate approach is for the user to post a commitment to some random value on his profile and then request
a credential from the issuer by supplying a proof of knowledge of the opening (committed random value) of the commitment.
The issuer is convinced that no one else could know the opening of the commitment which was posted by the user. Note that
the user is proving knowledge of the committed value and not revealing it to the issuer because revealing the value will
allow the issuer to request a similar credential from some another issuer of it the revealed value is leaked then someone
else can impersonate the user.
The test shows a complete example.
The code for composite proof lives here. See the tests here for various scenarios. For a more involved demo with multiple BBS signatures being used with accumulator and knowledge of signatures being proved before requesting blind signatures, see here. This test paints a picture where before getting any credential, a user has to prove possession of a credential and membership in an accumulator (except the 1st credential).
Note: This section assumes you have read some of the previous examples on composite proof.
A complete example as a test is here
Say a verifier wants the prover to encrypt an attribute from his credential for a 3rd party say a regulator. The verifier should be
able to check that the prover did encrypt a specific attribute from his credential and not some arbitrary value. Also, the verifier
should be able to check that the ciphertext is encrypted for the specific public key. This is achieved through verifiable
encryption and implemented using a protocol called SAVER.
For this, the decryptor needs to do a setup where it creates several parameters including encrytion key, decryption key,
SNARK proving key and verification key, etc. The decryptor then publishes the public parameters. In the snippet below,
snarkPk, encryptionKey, decryptionKey and gens are published.
import { SaverEncryptionGens } from '@docknetwork/crypto-wasm-ts';
const encGens = SaverEncryptionGens.generate();
const [snarkPk, secretKey, encryptionKey, decryptionKey] = SaverDecryptor.setup(encGens);
SaverDecryptor.setup above takes an optional parameter chunkBitSize which can make the encryption and proving faster (or slower)
while making decryption slower (or faster). Since encryption and proving are done more often, a higher default value of 16
is chosen for this parameter. Note that once parameters have been created with a certain value of chunkBitSize, the same value
should be used while encryption, decryption, proving and verification (as shown below).
For signers (issuers of credentials), it's important to encode attributes that need to be verifiably encrypted using a reversible
encoding as the decryption might happen much later than the proof verification and thus the decryptor should be able to independently
recover the actual attributes. This situation is different from selective disclosure where the actual attributes are given to the
verifier who can then encode the attributes before verifying the proof. One such pair of functions are Signature.reversibleEncodeStringForSigning
and Signature.reversibleDecodeStringForSigning and you can see its use in the above-mentioned test. Theese conversions are abstracted in this Encoders class and you can see the usage
in these tests of the Encoder initialized here.
For creating the proof of knowledge of the BBS signature and verifiably encrypting an attribute, the prover creates the following 2 statements.
import { SaverChunkedCommitmentKey } from '@docknetwork/crypto-wasm-ts';
// Signer's parameters
let sigParams: BBSSignatureParams, sigPk: BBSPublicKey, sig: BBSSignature;
// Signed messages
let messages: Uint8Array[];
...
...
// The value used by decryptor during setup
let chunkBitSize = ...;
...
...
// The following is either created by the verifier and is shared with the prover or created by the prover using a public bytes
// as argument to `SaverChunkedCommitmentKey.generate`
const gens = SaverChunkedCommitmentKey.generate(<some public bytes>);
...
...
// Uncompressed form of `gens` created above
const commKey = gens.decompress();
// Uncompressed form of other parameters created by decryptor
const saverEncGens = encGens.decompress();
const saverEk = encryptionKey.decompress();
const snarkProvingKey = snarkPk.decompress();
...
...
const statement1 = Statement.bbsSignatureProverConstantTime(sigParams, revealedMsgs, false);
const statement2 = Statement.saverProver(saverEncGens, commKey, saverEk, snarkProvingKey, chunkBitSize);
const proverStatements = new Statements();
proverStatements.add(statement1);
proverStatements.add(statement2);
statement1 is the for proving knowledge of a BBS signature as seen in previous examples. statement2 is for proving the encryption of message from a
BBS signature. Some things to note about this statement.
Statement.saverProver because it is being created by a prover. A verifier would have
used Statement.saverVerifier to create it and one of the arguments would be different (shown below).saverEncGens is the encryption generators created by decryptor. However, before they are passed to Statement.saverProver, the are uncompressed (ref. elliptic curve point compression) as shown in the above snippet. Uncompressing them doubles their size but makes them faster to work with. However, if you still want to use the compressed parameters use Statement.saverProverFromCompressedParamssaverEk is the encryption key created by the decryptor during setup but is uncompressed.snarkProvingKey is the proving key created by the decryptor during setup but is uncompressed.The prover then establishes the equality between the message in the BBS signature and the message being encrypted by using
WitnessEqualityMetaStatement as below. encMsgIdx is the index of the message being encrypted in the array of signed
messages under BBS, messages. For the second statement, there is only 1 witness, thus the index 0.
const witnessEq = new WitnessEqualityMetaStatement();
witnessEq.addWitnessRef(0, encMsgIdx);
witnessEq.addWitnessRef(1, 0);
const metaStatements = new MetaStatements();
metaStatements.add(MetaStatement.witnessEquality(witnessEq));
The prover then creates witness for both statements. The message messages[encMsgIdx] passed to Witness.saver is the
message being encrypted. unrevealedMsgs passed to Witness.bbsSignatureConstantTime is created from messages and consists of
messages not being revealed to the verifier.
const witness1 = Witness.bbsSignatureConstantTime(sig, unrevealedMsgs, false);
const witness2 = Witness.saver(messages[encMsgIdx]);
const witnesses = new Witnesses();
witnesses.add(witness1);
witnesses.add(witness2);
The prover then creates a proof specification using QuasiProofSpec. This is different from ProofSpec object seen in
previous examples as it does not call WASM to get a proof specification object and thus is more efficient.
Now prover creates the proof using CompositeProof.generateUsingQuasiProofSpec
import { QuasiProofSpec } from '@docknetwork/crypto-wasm-ts';
const proverProofSpec = new QuasiProofSpec(proverStatements, metaStatements);
const proof = CompositeProof.generateUsingQuasiProofSpec(proverProofSpec, witnesses);
Similarly, the verifier also creates 2 statements and the same meta statement to verify the proof.
// Get the uncompressed verifying key from the compressed proving key.
const snarkVerifyingKey = snarkPk.getVerifyingKeyUncompressed();
const statement1 = Statement.bbsSignatureProverConstantTime(sigParams, revealedMsgs, false);
const statement2 = Statement.saverVerifier(saverEncGens, commKey, saverEk, snarkVerifyingKey, chunkBitSize);
const verifierStatements = new Statements();
verifierStatements.add(statement1);
verifierStatements.add(statement2);
const witnessEq = new WitnessEqualityMetaStatement();
witnessEq.addWitnessRef(0, encMsgIdx);
witnessEq.addWitnessRef(1, 0);
const metaStatements = new MetaStatements();
metaStatements.add(MetaStatement.witnessEquality(witnessEq));
The above has a few differences from the prover's statements:
Statement.saverProver, verifier uses Statement.saverVerifier.The verifier now creates the proof specification and verifies the proof.
const verifierProofSpec = new QuasiProofSpec(verifierStatements, metaStatements);
// result.verified should be true for the proof to be valid.
const result = proof.verifyUsingQuasiProofSpec(verifierProofSpec);
The verifier will now extract the ciphertext from the proof so that it can share that with the decryptor later. Here 1
passed to proof.getSaverCiphertext is the index (0-based) of the statement in the list of statements being proven and the
statement from verifiable encryption was the 2nd one.
const ciphertext = proof.getSaverCiphertext(1);
The decryptor can decrypt the ciphertext to get message that was encrypted.
const saverDk = decryptionKey.decompress();
// decrypted.message is the message
const decrypted = SaverDecryptor.decryptCiphertext(ciphertext, saverSk, saverDk, snarkVerifyingKey, chunkBitSize);
Sometimes the verifier might want to know the decrypted message but might not trust that the decryptor to correctly tell him the decrypted message. In this it can verify the decryption done by the decryptor as below
// result.verified should be true
const result = ciphertext.verifyDecryption(decrypted, saverDk, snarkVerifyingKey, saverEncGens, chunkBitSize);
Note: This section assumes you have read some of the previous examples on composite proof.
A complete example as a test is here
Allows a verifier to check that some attribute of the credential satisfies given bounds min and max, i.e. min <= message < max
without learning the attribute itself. Both min and max are positive integers. This can be implemented using different protocols,
The above mentioned test uses all these variations.
To work with negative integers or decimal numbers, they must be converted to positive integers first and this conversion must happen before these are signed.
When working with negative integers, add the absolute value of the smallest (negative) integer to all values including bounds. Eg, if the smallest negative
number a value can be is -300, the signer should sign value + 300 to ensure that values are always positive. During the bound check, say the verifier has to
check if the value is between -200 and 50, the verifier should ask the prover to the bounds as 100 (-200 + 300) and 350 (50 + 300). When working with decimal
numbers, convert them to integers by multiplying with a number to make it integer, like if a decimal value can have maximum of 3 decimal places, they should be
multiplied by 1000. The test mentioned above shows these scenarios.
The conversions defined in the above tests are abstracted in this Encoders class and you can see the usage
in these tests of the Encoder initialized here.
For this, the verifier needs to first create the setup parameters which he then shares with the prover. Note that the
verifier does not have to create them each time a proof needs to be verified, but only once and publish them somewhere
such that all provers interacting with the proof can use them.
In the following snippet, the verifier asks to prove that a certain message satisfies the lower and upper bounds min and max,
i.e. min <= message < max. Note than both bounds are positive integers and lower bound is inclusive but upper bound is not.
To change from exclusive to inclusive bounds and vice-versa, add or subtract 1 from bounds. The snippet shows LegoGroth16.
import { BoundCheckSnarkSetup } from '@docknetwork/crypto-wasm-ts';
const provingKey = BoundCheckSnarkSetup();
For creating the proof of knowledge of the BBS signature and one of the signed message being in certain bounds, the prover creates the following 2 statements.
// Signer's parameters
let sigParams: BBSSignature, pk: BBSPublicKey, sig: BBSSignature;
// Signed messages - already encoded
let messages: Uint8Array[];
...
// define the min and max bounds
let min: number = ...;
let max: number = ...;
...
// Decompress the proving key
const snarkProvingKey = provingKey.decompress();
const statement1 = Statement.bbsSignatureProverConstantTime(sigParams, revealedMsgs, false);
const statement2 = Statement.boundCheckLegoProver(min, max, snarkProvingKey);
const proverStatements = new Statements();
proverStatements.add(statement1);
proverStatements.add(statement2);
statement1 is the for proving knowledge of a BBS signature as seen in previous examples. statement2 is for proving the bounds of a message from the BBS signature.
Some things to note about this statement:
Statement.boundCheckLegoProver because it is being created by a prover. A verifier would have
used Statement.boundCheckLegoVerifier to create it and one of the arguments would be different (shown below).snarkProvingKey is the public parameter created by the verifier. However, before they are passed to Statement.boundCheckLegoProver, they are uncompressed (ref. elliptic curve point compression) as shown in the above snippet. Uncompressing them doubles their size but makes them faster to work with. However, if you still want to use the compressed parameters use Statement.boundCheckLegoProverFromCompressedParamsThe prover then establishes the equality between the message in the BBS signature and the bounded message by using
WitnessEqualityMetaStatement as below. msgIdx is the index of the bounded message in the array of signed messages
under BBS, messages. For the second statement, there is only 1 witness, thus the index 0.
import { WitnessEqualityMetaStatement, MetaStatement, MetaStatements } from '@docknetwork/crypto-wasm-ts';
const witnessEq = new WitnessEqualityMetaStatement();
const msgIdx = 3; // the index of the SSN number
witnessEq.addWitnessRef(0, msgIdx);
witnessEq.addWitnessRef(1, 0);
const metaStatements = new MetaStatements();
metaStatements.add(MetaStatement.witnessEquality(witnessEq));
The prover then creates witness for both statements. The message messages[msgIdx] passed to Witness.boundCheckLegoGroth16 is the
bounded message. unrevealedMsgs passed to Witness.bbsSignatureConstantTime is created from messages and consists of
messages not being revealed to the verifier.
const witness1 = Witness.bbsSignatureConstantTime(sig, unrevealedMsgs, false);
const witness2 = Witness.boundCheckLegoGroth16(messages[msgIdx]);
const witnesses = new Witnesses();
witnesses.add(witness1);
witnesses.add(witness2);
The prover then creates a proof specification using QuasiProofSpec. This is different from ProofSpec object seen in
previous examples as it does not call WASM to get a proof specification object and thus is more efficient.
Now prover creates the proof using CompositeProof.generateUsingQuasiProofSpec
import { QuasiProofSpec } from '@docknetwork/crypto-wasm-ts';
const proverProofSpec = new QuasiProofSpec(proverStatements, metaStatements);
const proof = CompositeProof.generateUsingQuasiProofSpec(proverProofSpec, witnesses);
Similarly, the verifier also creates 2 statements and the same meta statement to verify the proof.
// Get the uncompressed verifying key from the compressed proving key.
const snarkVerifyingKey = provingKey.getVerifyingKeyUncompressed();
const statement1 = Statement.bbsSignatureVerifierConstantTime(sigParams, sigPk, revealedMsgs, false);
const statement2 = Statement.boundCheckLegoVerifier(min, max, snarkVerifyingKey);
const verifierStatements = new Statements();
verifierStatements.add(statement1);
verifierStatements.add(statement2);
const witnessEq = new WitnessEqualityMetaStatement();
witnessEq.addWitnessRef(0, msgIdx);
witnessEq.addWitnessRef(1, 0);
const metaStatements = new MetaStatements();
metaStatements.add(MetaStatement.witnessEquality(witnessEq));
The above has a few differences from the prover's statements:
Statement.boundCheckLegoProver, verifier uses Statement.boundCheckLegoVerifier.The verifier now creates the proof specification and verifies the proof.
const verifierProofSpec = new QuasiProofSpec(verifierStatements, metaStatements);
// result.verified should be true for the proof to be valid.
const result = proof.verifyUsingQuasiProofSpec(verifierProofSpec);
You might notice some public parameters are huge and also the statements involving them take noticeable time to create. Eg,
snarkProvingKey, snarkVerifyingKey, saverEk are huge and thus creating Statement.saverProver, Statement.saverVerifier,
Statement.boundCheckLegoProver and Statement.boundCheckLegoVerifier take some time to create. This becomes a bigger problem
when several messages need to be encrypted for the same decryptor or bounds over several messages need to be proved.
To solve this, the public parameters don't need to be passed directly to the Statements. They can be wrapped in a SetupParam
and then a reference to them is passed as an argument in place of the parameter itself to the Statement. See the snippet
below for creating 2 statements for verifiable encryption for the same setup parameters:
import { SetupParam } from '@docknetwork/crypto-wasm-ts';
// Prover creates an array of `SetupParam`s
const proverSetupParams = [];
proverSetupParams.push(SetupParam.saverEncryptionGensUncompressed(saverEncGens));
proverSetupParams.push(SetupParam.saverCommitmentGensUncompressed(commKey));
proverSetupParams.push(SetupParam.saverEncryptionKeyUncompressed(saverEk));
proverSetupParams.push(SetupParam.saverProvingKeyUncompressed(snarkProvingKey));
// Passing reference to parameters as array indices from `proverSetupParams`
const statement3 = Statement.saverProverFromSetupParamRefs(0, 1, 2, 3, chunkBitSize);
const statement4 = Statement.saverProverFromSetupParamRefs(0, 1, 2, 3, chunkBitSize);
Note the use of Statement.saverProverFromSetupParamRefs rather than Statement.saverProver. The arguments:
proverSetupParamsproverSetupParamsproverSetupParamsproverSetupParamsNow the prover creates the proof specification by passing SetupParams array as well.
const proverStatements = new Statements();
...
proverStatements.add(statement3);
proverStatements.add(statement4);
...
...
...
const proverProofSpec = new QuasiProofSpec(proverStatements, metaStatements, proverSetupParams);
const proof = CompositeProof.generateUsingQuasiProofSpec(proverProofSpec, witnesses);
Similarly, the verifier can create his own SetupParams array for his proof specification and then proof
const verifierSetupParams = [];
verifierSetupParams.push(SetupParam.saverEncryptionGensUncompressed(saverEncGens));
verifierSetupParams.push(SetupParam.saverCommitmentGensUncompressed(commKey));
verifierSetupParams.push(SetupParam.saverEncryptionKeyUncompressed(saverEk));
verifierSetupParams.push(SetupParam.saverVerifyingKeyUncompressed(snarkVerifyingKey));
const statement5 = Statement.saverVerifierFromSetupParamRefs(0, 1, 2, 3, chunkBitSize);
const statement6 = Statement.saverVerifierFromSetupParamRefs(0, 1, 2, 3, chunkBitSize);
const verifierStatements = new Statements();
...
...
verifierStatements.add(statement5);
verifierStatements.add(statement6);
const verifierProofSpec = new QuasiProofSpec(verifierStatements, metaStatements, verifierSetupParams);
const result = proof.verifyUsingQuasiProofSpec(verifierProofSpec);
For a complete example, see these tests
Similarly, for bound checks, use Statement.boundCheckLegoProverFromSetupParamRefs and Statement.boundCheckVerifierFromSetupParamRefs.
For complete example, see these tests
The above interfaces have been found to be a bit difficult to work with when signing messages/credentials that are represented as JS objects like
{
"fname": "John",
"lname": "Smith",
"sensitive": {
"secret": "my-secret-that-wont-tell-anyone",
"email": "john.smith@example.com",
"SSN": "123-456789-0",
"user-id": "user:123-xyz-#"
},
"location": {
"country": "USA",
"city": "New York"
},
"timeOfBirth": 1662010849619,
"physical": {
"height": 181.5,
"weight": 210,
"BMI": 23.25
},
"score": -13.5
}
Here are some utilities to make this task a bit easier. The idea is to flatten the JSON, sort the keys alphabetically
to have a list with deterministic order and then use the encoder to encode each value as a field element (a number between 0 and another large number).
The encoder can be configured to use different encoding functions for different keys to convert values from different types
like string, positive or negative integers or decimal numbers to field elements.
The tests here contain plenty of examples.
Simple predicates like a range proof or equality of messages in zero knowledge are already hardcoded in the library but we cannot imagine all the possible predicates different use-cases can require. We expect developers to write these predicates in a programming language that we can then use to create zero-knowledge proofs. We currently support Circom, version 2. The predicates can be written as Circom programs and then compiled for curve BLS12-381. The generated R1CS and WASM can then be feed to the composite proof system to generate a zero knowledge proof of the predicate.
The workflow is this:
The steps 1-3 are done by the verifier and the result of these steps, i.e. the program (.circom file), R1CS (.r1cs file),
WASM (.wasm file), proving and verification key are shared with any potential prover (published or shared P2P). Step 4 is
done by the prover and step 5 again by the verifier. Since R1CS and WASM files are harder to inspect that Circom programs,
to guard against a verifier tricking the prover to prove unintended predicates (and thus reveal more information than required),
a prover can take the Circom program and generate the R1CS and WASM files himself (do step 2 as well).
See some of the following tests for Circom usage:
The Circom programs and corresponding R1CS and WASM files for the tests are here.
The composite proof system is used to implement anonymous credentials. See here for details.
Slides and video for a presentation given at SSI meetup. Most of the presentation goes over the code, mostly anonymous credentials from this library.