An easy, lightweight, and secure module for fetching JWK Public Keys and verifying JWTs. Great for OIDC, Auth0, JWKs URLs.
Go to file
AJ ONeal 6730e1321e bugfix: properly pass details and msg 2021-10-21 13:40:12 -06:00
lib bugfix: properly pass details and msg 2021-10-21 13:40:12 -06:00
.gitignore v1.1.8: cleanup and support forced keys for verification 2019-03-09 02:50:14 -07:00
.jshintrc chore: add linter config 2021-10-20 18:08:27 -06:00
.prettierrc.json rename .prettierrc.json 2021-06-15 16:15:43 -06:00
CHANGELOG.md feature: add additional (standardized) error messages 2021-10-21 13:21:41 -06:00
LICENSE v1.1.10: update package.json and LICENSE 2019-03-13 17:32:35 -06:00
README.md feature: add additional (standardized) error messages 2021-10-21 13:21:41 -06:00
keyfetch-test.js security warning -> issuer error, async/await 2021-06-15 17:22:38 -06:00
keyfetch.js feature: add additional (standardized) error messages 2021-10-21 13:21:41 -06:00
package-lock.json 2.1.0 2021-10-21 13:22:08 -06:00
package.json 2.1.0 2021-10-21 13:22:08 -06:00

README.md

keyfetch

Lightweight support for fetching JWKs.

Fetches JSON native JWKs and exposes them as PEMs that can be consumed by the jsonwebtoken package (and node's native RSA and ECDSA crypto APIs).

Features

Works great for

  • jsonwebtoken (Auth0)
  • OIDC (OpenID Connect)
  • .well-known/jwks.json (Auth0, Okta)
  • Other JWKs URLs

Crypto Support

  • JWT verification
  • RSA (all variants)
  • EC / ECDSA (NIST variants P-256, P-384)
  • Sane error codes
  • esoteric variants (excluded to keep the code featherweight and secure)

Table of Contents

  • Install
  • Usage
  • API
    • Auth0 / Okta
    • OIDC
  • Errors
  • Change Log

Install

npm install --save keyfetch

Usage

Retrieve a key list of keys:

var keyfetch = require("keyfetch");

keyfetch.oidcJwks("https://example.com/").then(function (results) {
    results.forEach(function (result) {
        console.log(result.jwk);
        console.log(result.thumprint);
        console.log(result.pem);
    });
});

Quick JWT verification (for authentication):

var keyfetch = require("keyfetch");
var jwt = "...";

keyfetch.jwt.verify(jwt).then(function (decoded) {
    console.log(decoded);
});

JWT verification (for authorization):

var options = { issuers: ["https://example.com/"], claims: { role: "admin" } };
keyfetch.jwt.verify(jwt, options).then(function (decoded) {
    console.log(decoded);
});

Verify a JWT with jsonwebtoken:

var keyfetch = require("keyfetch");
var jwt = require("jsonwebtoken");
var auth = "..."; // some JWT
var token = jwt.decode(auth, { json: true, complete: true });

if (!isTrustedIssuer(token.payload.iss)) {
    throw new Error("untrusted issuer");
}

keyfetch.oidcJwk(token.header.kid, token.payload.iss).then(function (result) {
    console.log(result.jwk);
    console.log(result.thumprint);
    console.log(result.pem);

    jwt.jwt.verify(jwt, { jwk: result.jwk });
});

Note: You might implement isTrustedIssuer one of these:

function isTrustedIssuer(iss) {
    return -1 !== ["https://partner.com/", "https://auth0.com/"].indexOf(iss);
}
function isTrustedIssuer(iss) {
    return (
        /^https:/.test(iss) && /(\.|^)example\.com$/.test(iss) // must be a secure domain
    ); // can be example.com or any subdomain
}

API

All API calls will return the RFC standard JWK SHA256 thumbprint as well as a PEM version of the key.

Note: When specifying id, it may be either kid (as in token.header.kid) or thumbprint (as in result.thumbprint).

JWKs URLs

Retrieves keys from a URL such as https://example.com/jwks/ with the format { keys: [ { kid, kty, exp, ... } ] } and returns the array of keys (as well as thumbprint and jwk-to-pem).

keyfetch.jwks(jwksUrl);
// Promises [ { jwk, thumbprint, pem } ] or fails
keyfetch.jwk(id, jwksUrl);
// Promises { jwk, thumbprint, pem } or fails

Auth0

If https://example.com/ is used as issuerUrl it will resolve to https://example.com/.well-known/jwks.json and return the keys.

keyfetch.wellKnownJwks(issuerUrl);
// Promises [ { jwk, thumbprint, pem } ] or fails
keyfetch.wellKnownJwk(id, issuerUrl);
// Promises { jwk, thumbprint, pem } or fails

OIDC

If https://example.com/ is used as issuerUrl then it will first resolve to https://example.com/.well-known/openid-configuration and then follow jwks_uri to return the keys.

keyfetch.oidcJwks(issuerUrl);
// Promises [ { jwk, thumbprint, pem } ] or fails
keyfetch.oidcJwk(id, issuerUrl);
// Promises { jwk, thumbprint, pem } or fails

Verify JWT

This can accept a JWT string (compact JWS) or a decoded JWT object (JWS).

This can be used purely for verifying pure authentication tokens, as well as authorization tokens.

keyfetch.jwt.verify(jwt, { strategy: "oidc" }).then(function (verified) {
    /*
    { protected: '...'  // base64 header
    , payload: '...'    // base64 payload
    , signature: '...'  // base64 signature
    , header: {...}     // decoded header
    , claims: {...}     // decoded payload
    }
  */
});

When used for authorization, it's important to specify a limited set of trusted issuers.
When using for federated authentication you may set issuers = ["*"] - but DO NOT trust claims such as email and email_verified.

If your authorization claims can be expressed as exact string matches, you can specify those too.

keyfetch.jwt.verify(jwt, {
  strategy: 'oidc',
  issuers: [ 'https://example.com/' ],
  //iss: 'https://example.com/',
  claims: { role: 'admin', sub: 'abc', group: 'xyz' }
}).then(function (verified) {
  • strategy may be oidc (default) , auth0, or a direct JWKs url.
  • issuers must be a list of https urls (though http is allowed for things like Docker swarm), or '*'
  • iss is like issuers, but only one
  • claims is an object with arbitrary keys (i.e. everything except for the standard iat, exp, jti, etc)
  • exp may be set to false if you're validating on your own (i.e. allowing time drift leeway)
  • jwks can be used to specify a list of allowed public key rather than fetching them (i.e. for offline unit tests)
  • jwk same as above, but a single key rather than a list

Decode JWT

try {
  console.log( keyfetch.jwt.decode(jwt) );
} catch(e) {
  console.error(e);
}
{ protected: '...'  // base64 header
, payload: '...'    // base64 payload
, signature: '...'  // base64 signature
, header: {...}     // decoded header
, claims: {...}     // decoded payload

It's easier just to show the code than to explain the example.

keyfetch.jwt.decode = function (jwt) {
    // Unpack JWS from "compact" form
    var parts = jwt.split(".");
    var obj = {
        protected: parts[0],
        payload: parts[1],
        signature: parts[2]
    };

    // Decode JWT properties from JWS as unordered objects
    obj.header = JSON.parse(Buffer.from(obj.protected, "base64"));
    obj.claims = JSON.parse(Buffer.from(obj.payload, "base64"));

    return obj;
};

Cache Settings

keyfetch.init({
    // set all keys at least 1 hour (regardless of jwk.exp)
    mincache: 1 * 60 * 60,

    // expire each key after 3 days (regardless of jwk.exp)
    maxcache: 3 * 24 * 60 * 60,

    // re-fetch a key up to 15 minutes before it expires (only if used)
    staletime: 15 * 60
});

There is no background task to cleanup expired keys as of yet. For now you can limit the number of keys fetched by having a simple whitelist.

Errors

JSON.stringify()d errors look like this:

{
  code: "INVALID_JWT",
  status: 401,
  details: [ "jwt.claims.exp = 1634804500", "DEBUG: helpful message" ]
  message: "token's 'exp' has passed or could not parsed: 1634804500"
}

SemVer Compatibility:

  • code & status will remain the same.
  • The message property of an error is NOT included in the semver compatibility guarantee (we intend to make them more client-friendly), neither is detail at this time (but it will be once we decide on what it should be).

For backwards compatibility with v1, the non-stringified message is the same as what it was in v1 (and the v2 message is client_message, which replaces message in v3). Don't rely on it. Rely on code.

Hint Code Status Message (truncated)
(developer error) DEVELOPER_ERROR 500 test...
(bad gateway) BAD_GATEWAY 502 The token could not be verified because our s...
(insecure issuer) MALFORMED_JWT 400 'test' is NOT secure. Set env 'KEYFETCH_ALLOW...
(parse error) MALFORMED_JWT 400 could not parse jwt: 'test'...
(no issuer) MALFORMED_JWT 400 'iss' is not defined...
(malformed exp) MALFORMED_JWT 400 token's 'exp' has passed or could not parsed:...
(expired) INVALID_JWT 401 token's 'exp' has passed or could not parsed:...
(inactive) INVALID_JWT 401 token's 'nbf' has not been reached or could n...
(bad signature) INVALID_JWT 401 token signature verification was unsuccessful...
(jwk not found old) INVALID_JWT 401 Retrieved a list of keys, but none of them ma...
(jwk not found) INVALID_JWT 401 No JWK found by kid or thumbprint 'test'...
(no jwkws uri) INVALID_JWT 401 Failed to retrieve openid configuration...
(unknown issuer) INVALID_JWT 401 token was issued by an untrusted issuer: 'tes...
(failed claims) INVALID_JWT 401 token did not match on one or more authorizat...

Change Log

Minor Breaking changes (with a major version bump):

  • v3.0.0 - reworked error messages (also available in v2.1.0 as client_message)
  • v2.0.0 - changes from the default issuers = ["*"] to requiring that an issuer (or public jwk for verification) is specified

See other changes in CHANGELOG.md.