Nostr: npub1mgvlrnf5hm9yf0n5mf9nqmvarhvxkc6remu5ec3vf8r0txqkuk7su0e7q2
Pedersen commitment range proofs on secp256k1.
Prove a value is within a range without revealing it.
- Age-gating — prove a user is 18+ or between 13 and 17 without revealing their birth date
- Income brackets — prove income is above a threshold for a loan without revealing the amount
- Credit scoring — prove a credit score is above a threshold without revealing the score
- Salary bands — prove a salary falls within a negotiating band without disclosing it
npm install @forgesworn/range-proofRequirements:
- ESM-only. There is no CommonJS build; use
import, notrequire. - Node.js 20.19 or later. This is the floor declared by the runtime dependencies
@noble/curvesand@noble/hashes(enginesinpackage.json). - Synchronous API. Every function returns its result directly. Nothing returns a Promise, so there is nothing to
await.
Each snippet below is self-contained: it imports what it uses and defines every variable.
import { createRangeProof, verifyRangeProof } from '@forgesworn/range-proof';
const value = 25; // the secret value, known only to the prover
const min = 18;
const max = 150;
// Prove that `value` is in [min, max] without revealing `value`
const proof = createRangeProof(value, min, max);
// Verifiers must supply the public range they expect
const valid = verifyRangeProof(proof, min, max); // truecreateRangeProof returns a RangeProof. verifyRangeProof returns a boolean.
Limits: value, min and max must be safe integers, min must be non-negative, max must be at least min, and value must lie in [min, max]. The width of the range, max - min, may be at most 2^32 - 1 (see Cryptography).
import { createAgeRangeProof, verifyAgeRangeProof } from '@forgesworn/range-proof';
// Prove age is between 8 and 12 (e.g. child category)
const childProof = createAgeRangeProof(10, '8-12');
const childValid = verifyAgeRangeProof(childProof, '8-12'); // true
// Prove age is between 13 and 17
const teenProof = createAgeRangeProof(15, '13-17');
const teenValid = verifyAgeRangeProof(teenProof, '13-17'); // true
// Prove age is 18 or over
const adultProof = createAgeRangeProof(25, '18+');
const adultValid = verifyAgeRangeProof(adultProof, '18+'); // trueThe age range string is one of two forms, digits only:
| Form | Example | Range proved |
|---|---|---|
'min-max' |
'8-12', '13-17' |
[min, max], both inclusive |
'min+' |
'18+' |
[min, 150]; the upper bound is fixed at 150 |
No spaces, signs, decimals or other characters are accepted; leading zeros are ('08-12' means 8 to 12). A malformed string is handled differently by the two functions:
createAgeRangeProofthrowsValidationError(Invalid age range format (expected "min-max" or "min+")).verifyAgeRangeProofnever throws and returnsfalse.
A well-formed string whose bounds are inverted, such as '12-8' or '151+', is also rejected: createAgeRangeProof throws ValidationError (Maximum must be >= minimum) and verifyAgeRangeProof returns false. An age outside the range, such as 17 for '18+', makes createAgeRangeProof throw ValidationError.
The age helpers are thin wrappers over the range functions. A proof from createAgeRangeProof(25, '18+') is the same as createRangeProof(25, 18, 150), and verifyRangeProof(proof, 18, 150) accepts it. The optional third argument, subjectPubkey, is the binding context described below.
Pass an optional context string to bind the proof to a specific credential or identity. A proof created with one context will not verify under a different context, preventing transplant attacks. The context may be any string up to 1024 UTF-8 bytes (longer strings throw ValidationError); an empty string is treated as no context.
import { createRangeProof, verifyRangeProof } from '@forgesworn/range-proof';
const value = 25;
const min = 18;
const max = 150;
const subjectPubkey = 'subject-pubkey-hex';
const proof = createRangeProof(value, min, max, subjectPubkey);
verifyRangeProof(proof, min, max, subjectPubkey); // true
verifyRangeProof(proof, min, max, 'other-subject'); // false: wrong context
verifyRangeProof(proof, min, max); // false: context missingimport { commit, verifyCommitment } from '@forgesworn/range-proof';
const c = commit(42);
// c.commitment: the public commitment point (66-character compressed hex, 02/03 prefix)
// c.blinding: the secret blinding factor (64-character lowercase hex)
// c.value: the committed value (kept secret)
// Open the commitment to verify
const valid = verifyCommitment(c.commitment, 42, c.blinding); // true
// Optionally supply your own blinding factor
const fixed = commit(42, 123456789n);The two functions take the blinding factor in different forms:
commit(value, blinding?)takes an optionalbigintin[1, N-1], whereNis the secp256k1 group order. A random factor is generated if you omit it. A non-bigint,0n, a negative value or a value ofNor more throwsValidationError. The returnedblindingis a 64-character zero-padded lowercase hex string.verifyCommitment(commitment, value, blinding)takesblindingas a hex string of 1 to 64 hex digits (any case, no0xprefix) that encodes a value belowN. It returnsfalsefor a wrong opening, malformed hex or an invalid point, and throwsValidationErroronly when an argument has the wrong JavaScript type (commitmentorblindingnot a string,valuenot a non-negative safe integer).
commit and createRangeProof are independent. createRangeProof makes its own commitment and keeps the blinding factor internal, so proof.commitment cannot be opened with verifyCommitment.
import {
createRangeProof,
serializeRangeProof,
deserializeRangeProof,
verifyRangeProof,
} from '@forgesworn/range-proof';
const proof = createRangeProof(25, 18, 150);
const json = serializeRangeProof(proof); // string
const proof2 = deserializeRangeProof(json); // RangeProof; throws ValidationError if malformed
const valid = verifyRangeProof(proof2, 18, 150); // truefunction createRangeProof(value: number, min: number, max: number, bindingContext?: string): RangeProof
function verifyRangeProof(proof: RangeProof, expectedMin: number, expectedMax: number, expectedBindingContext?: string): boolean
function createAgeRangeProof(age: number, ageRange: string, subjectPubkey?: string): RangeProof
function verifyAgeRangeProof(proof: RangeProof, expectedAgeRange: string, expectedSubjectPubkey?: string): boolean
function commit(value: number, blinding?: bigint): PedersenCommitment
function verifyCommitment(commitment: string, value: number, blinding: string): boolean
function serializeRangeProof(proof: RangeProof): string
function deserializeRangeProof(json: string): RangeProofAll of these are synchronous. The package also exports the error classes RangeProofError, ValidationError and CryptoError (see Error Handling). llms-full.txt has the full reference.
RangeProof and PedersenCommitment are exported as types:
interface RangeProof {
commitment: string; // overall commitment C = v*G + r*H (compressed hex)
min: number; // minimum of the range (public)
max: number; // maximum of the range (public)
bits: number; // bits used: ceil(log2(max - min + 1))
lowerProof: BitProof[]; // bit proofs for (value - min)
upperProof: BitProof[]; // bit proofs for (max - value)
lowerCommitment: string; // commitment to (value - min)
upperCommitment: string; // commitment to (max - value)
sumBindingE: string; // sum-binding Schnorr challenge (hex scalar)
sumBindingS: string; // sum-binding Schnorr response (hex scalar)
commitBindingE: string; // commitment-binding Schnorr challenge (hex scalar)
commitBindingS: string; // commitment-binding Schnorr response (hex scalar)
context?: string; // binding context, present only if one was supplied
}
// Not exported by name; every field is a hex string
interface BitProof {
commitment: string;
e0: string;
s0: string;
e1: string;
s1: string;
}
interface PedersenCommitment {
commitment: string; // C = v*G + r*H (66-character compressed hex), safe to share
blinding: string; // the blinding factor r (64-character hex), keep secret
value: number; // the committed value, keep secret
}A RangeProof contains no blinding factor and no secret value, so it can be sent to a verifier as is (or via serializeRangeProof).
Three error classes, all importable from the package:
import {
RangeProofError, // base class
ValidationError, // malformed inputs, out-of-range values, bad JSON
CryptoError, // range width too large, cryptographic failures
} from '@forgesworn/range-proof';createRangeProof throws on invalid inputs:
import { createRangeProof, ValidationError, CryptoError } from '@forgesworn/range-proof';
const value = 25;
const min = 18;
const max = 150;
const bindingContext = 'subject-pubkey-hex';
try {
const proof = createRangeProof(value, min, max, bindingContext);
} catch (err) {
if (err instanceof ValidationError) {
// 'Range proof values must be safe integers'
// 'Minimum must be non-negative'
// 'Maximum must be >= minimum'
// 'Value is not within the specified range'
// 'Binding context exceeds maximum length (1024 bytes)'
}
if (err instanceof CryptoError) {
// 'Range too large for range proof (max 2^32)': max - min is 2^32 or more
}
}createAgeRangeProof throws ValidationError for a malformed age range string (see
Age range proofs) and for everything createRangeProof rejects.
commit throws ValidationError for a negative or non-integer value and for an invalid
blinding factor. verifyCommitment returns false for a wrong opening, bad hex or an invalid
point, and throws ValidationError only for arguments of the wrong type.
verifyRangeProof never throws — it returns false for any invalid or tampered
proof. This is a deliberate design choice: verification is a boolean question.
import { createRangeProof, verifyRangeProof } from '@forgesworn/range-proof';
const proof = createRangeProof(25, 18, 150);
const valid = verifyRangeProof(proof, 18, 150);
// valid is true or false, never an exceptionverifyAgeRangeProof also never throws; a malformed age range string gives false.
The commitment is taken from the proof itself. Verification answers "does this
proof demonstrate a value in [min, max] for this commitment?" — it does not
attest to whose commitment it is. Callers must independently anchor
proof.commitment to externally known state (e.g. a commitment the prover
previously published or that is bound to their identity). The expected range
and binding context passed to verifyRangeProof must come from the verifier's
own policy, not from the proof.
deserializeRangeProof throws ValidationError for malformed JSON, missing fields,
or invalid hex values. This is where you should handle errors when loading proofs from
untrusted sources:
import {
deserializeRangeProof,
verifyRangeProof,
ValidationError,
createRangeProof,
serializeRangeProof,
} from '@forgesworn/range-proof';
// Full verification pipeline with error handling
function verifyProofFromJson(
json: string,
expectedMin: number,
expectedMax: number,
expectedContext?: string,
): boolean {
try {
const proof = deserializeRangeProof(json);
return verifyRangeProof(proof, expectedMin, expectedMax, expectedContext);
} catch (err) {
if (err instanceof ValidationError) {
// Malformed proof data: reject
console.error('Invalid proof format:', err.message);
return false;
}
throw err; // unexpected error, re-throw
}
}
const json = serializeRangeProof(createRangeProof(25, 18, 150));
const ok = verifyProofFromJson(json, 18, 150); // true- Pedersen commitments:
C = v*G + r*HwhereHis a nothing-up-my-sleeve second generator derived by hashing'secp256k1-pedersen-H-v1'to a curve point. - Bit-decomposition range proofs: CDS OR-composition proving each bit is 0 or 1, with sum-binding and commitment-binding Schnorr proofs tying the bits to the range constraint and public commitment.
- Fiat-Shamir: domain-separated with
'pedersen-bit-proof-v1','pedersen-sum-binding-v1', and'pedersen-commitment-binding-v1'. - Maximum range width:
max - minmay be at most 2^32 - 1, so a proof covers at most 2^32 distinct values (32 bits per side).createRangeProofthrowsCryptoErrorbeyond that. The limit applies only to the widthmax - min:value,minandmaxthemselves can be any non-negative safe integers (for examplecreateRangeProof(2 ** 40 + 3, 2 ** 40, 2 ** 40 + 10)is fine).
The second generator H is critical to Pedersen commitment security. Nobody must know
log_G(H) — if they did, they could open a commitment to any value. H is derived
deterministically using a nothing-up-my-sleeve construction:
Algorithm: try-and-increment hash-to-point
1. seed = UTF-8 bytes of 'secp256k1-pedersen-H-v1' (23 bytes)
2. For counter i = 0, 1, 2, ... up to 255:
a. buf = seed || byte(i) (24 bytes)
b. h = SHA-256(buf) (32 bytes)
c. candidate = 0x02 || h (33 bytes — compressed point, even Y)
d. If candidate is a valid secp256k1 point → H = candidate; stop
e. Otherwise → increment i and retry
3. If no valid point found in 256 iterations → throw CryptoError
In practice, counter i = 0 produces a valid point on the first try. The algorithm is
deterministic — every implementation produces the same H from the same seed string.
The security property: H is derived entirely from a fixed ASCII string with no
trapdoor. The hash acts as a random oracle, and nobody can compute log_G(H) without
breaking the discrete logarithm assumption on secp256k1.
This is the same "hash-and-pray" technique used by Bulletproofs (Bünz et al. 2018) and other Pedersen-based protocols where a second generator is needed without a trusted setup.
ForgeSworn builds open-source cryptographic identity, payments, and coordination tools for Nostr.
| Library | What it does |
|---|---|
| nsec-tree | Deterministic sub-identity derivation |
| ring-sig | SAG/LSAG ring signatures on secp256k1 |
| range-proof | Pedersen commitment range proofs |
| canary-kit | Coercion-resistant spoken verification |
| spoken-token | Human-speakable verification tokens |
| toll-booth | L402 payment middleware |
| geohash-kit | Geohash toolkit with polygon coverage |
| nostr-attestations | NIP-VA verifiable attestations |
| dominion | Epoch-based encrypted access control |
| nostr-veil | Privacy-preserving Web of Trust |
MIT