openid-client-kingfisher

Disclamer: this work is a fork of the excellent package openid-client, with some changes made to ensure compatibility with NHS Identity.

openid-client-kingfisher is a server side OpenID Relying Party (RP, Client) implementation for Node.js, supports passport.

Notice: openid-client ^2.0.x drops support for Node.js versions less than lts/boron(6.9.0) due to Node.js lts/argon end of life on 2018-04-30. See the CHANGELOG for a complete list of deprecations and changes.

Table of Contents

• Implemented specs & features
• Certification
• Get started
• Usage
• Usage with passport
• Configuration

Implemented specs & features

The following client/RP features from OpenID Connect/OAuth2.0 specifications are implemented by openid-client.

• OpenID Connect Core 1.0
  • Authorization Callback
    • Authorization Code Flow
    • Implicit Flow
    • Hybrid Flow
  • UserInfo Request
  • Fetching Distributed Claims
  • Unpacking Aggregated Claims
  • Offline Access / Refresh Token Grant
  • Client Credentials Grant
  • Password Grant
  • Client Authentication
    • none
    • client_secret_basic
    • client_secret_post
    • client_secret_jwt
    • private_key_jwt
• RFC8414 - OAuth 2.0 Authorization Server Metadata and OpenID Connect Discovery 1.0
  • Discovery of OpenID Provider (Issuer) Metadata
  • Discovery of OpenID Provider (Issuer) Metadata via user provided inputs (see WebFinger)
• OpenID Connect Dynamic Client Registration 1.0
  • Dynamic Client Registration request
  • Client initialization via registration client uri
• RFC7009 - OAuth 2.0 Token revocation
  • Client Authenticated request to token revocation
• RFC7662 - OAuth 2.0 Token introspection
  • Client Authenticated request to token introspection

Updates to features defined in draft or experimental specifications are released as MINOR library versions, if you utilize these consider using the tilde ~ operator in your package.json since breaking changes may be introduced as part of these specification updates.

Certification

Filip Skokan has certified that openid-client conforms to the RP Basic, RP Implicit, RP Hybrid, RP Config, RP Dynamic and RP Form Post profiles of the OpenID Connect™ protocol.

If you want to quickly add OpenID Connect authentication to Node.js apps, feel free to check out Auth0's Node.js SDK and free plan at auth0.com/overview.

Get started

On the off-chance you want to manage multiple clients for multiple issuers you need to first get an Issuer instance.

via Discovery (recommended)

const { Issuer } = require('openid-client');
Issuer.discover('https://accounts.google.com') // => Promise
  .then(function (googleIssuer) {
    console.log('Discovered issuer %s %O', googleIssuer.issuer, googleIssuer.metadata);
  });

manually

const { Issuer } = require('openid-client');
const googleIssuer = new Issuer({
  issuer: 'https://accounts.google.com',
  authorization_endpoint: 'https://accounts.google.com/o/oauth2/v2/auth',
  token_endpoint: 'https://www.googleapis.com/oauth2/v4/token',
  userinfo_endpoint: 'https://www.googleapis.com/oauth2/v3/userinfo',
  jwks_uri: 'https://www.googleapis.com/oauth2/v3/certs',
}); // => Issuer
console.log('Set up issuer %s %O', googleIssuer.issuer, googleIssuer.metadata);

Now you can create your Client.

manually (recommended)

You should provide at least the following metadata: client_id, client_secret, id_token_signed_response_alg (defaults to RS256) and token_endpoint_auth_method (defaults to client_secret_basic) for a basic client definition, but you may provide any IANA registered client metadata.

const client = new googleIssuer.Client({
  client_id: 'zELcpfANLqY7Oqas',
  client_secret: 'TQV5U29k1gHibH5bx1layBo0OSAvAbRT3UYW3EWrSYBB5swxjVfWUa1BS8lqzxG/0v9wruMcrGadany3'
}, [keystore]); // => Client

keystore is an optional argument for instantiating a client with configured asymmetrical ID Token or UserInfo response encryption.

via registration client uri

Should your oidc provider have provided you with a registration client uri and registration access token you can also have the Client discovered.

googleIssuer.Client.fromUri(registration_client_uri, registration_access_token, [keystore]) // => Promise
  .then(function (client) {
    console.log('Discovered client %s %O', client.client_id, client.metadata);
  });

keystore is an optional argument for instantiating a client through registration client uri with configured asymmetrical ID Token or UserInfo response encryption.

Usage

Getting authorization url

client.authorizationUrl({
  redirect_uri: 'https://client.example.com/callback',
  scope: 'openid email',
}); // => String (URL)

You can also get HTML body of a self-submitting form to utilize POST to the authorization url with #authorizationPost method, same signature as #authorizationUrl.

client.authorizationPost({
  redirect_uri: 'https://client.example.com/callback',
  scope: 'openid email',
}); // => String (Valid HTML body)

Processing callback

const { state, response_type } = session[authorizationRequestState];
client.authorizationCallback('https://client.example.com/callback', request.query, { state, response_type }) // => Promise
  .then(function (tokenSet) {
    console.log('received and validated tokens %j', tokenSet);
    console.log('validated id_token claims %j', tokenSet.claims);
  });

Aside from state and response_type, checks for nonce (implicit and hybrid responses) and max_age are implemented. id_token signature and claims validation does not need to be requested, it is done automatically.

OP Errors - OpenIdConnectError

When the OpenID Provider returns an OIDC formatted error from either authorization callbacks or any of the JSON responses the library will reject a given Promise with OpenIdConnectError instance.

The message of this error is "${error} (${error_description})". However the OpenIdConnectError object also has the following properties:

• error
• error_description
• error_uri
• state
• scope

Values are undefined if these were not provided in the response. Additionally, for API call responses a response property is available with the response object from the used http client.

Handling multiple response modes

When handling multiple response modes with one single pass you can use #callbackParams to get the params object from the koa/express/node request object or a url string. (http.IncomingMessage). If form_post is your response_type you need to include a body parser prior.

client.callbackParams('https://client.example.com/cb?code=code'); // => { code: 'code' };
client.callbackParams('/cb?code=code'); // => { code: 'code' };

// example koa v2.x w/ koa-body
app.use(bodyParser({ patchNode: true }));
app.use(async function (ctx, next) {
  const params = client.callbackParams(ctx.request.req); // => parsed url query or body object
  // ...
});

// example express w/ bodyParser
app.use(bodyParser.urlencoded({ extended: false }));
app.use(function (req, res, next) {
  const params = client.callbackParams(req); // => parsed url query or body object
  // ...
});

Refreshing a token

client.refresh(refreshToken) // => Promise
  .then(function (tokenSet) {
    console.log('refreshed and validated tokens %j', tokenSet);
    console.log('refreshed id_token claims %j', tokenSet.claims);
  });

Tip: accepts TokenSet as well as direct refresh token values;

Revoke a token

client.revoke(token, [tokenTypeHint]) // => Promise
  .then(function (response) {
    console.log('revoked token %s', token, response);
  });

Introspect a token

client.introspect(token, [tokenTypeHint]) // => Promise
  .then(function (response) {
    console.log('token details %j', response);
  });

Fetching userinfo

client.userinfo(accessToken) // => Promise
  .then(function (userinfo) {
    console.log('userinfo %j', userinfo);
  });

Tip: accepts TokenSet as well as direct access token values;

via POST

client.userinfo(accessToken, { verb: 'post' }); // => Promise

with extra query/body payload

client.userinfo(accessToken, { params: { fields: 'email,ids_for_business' } }); // => Promise

auth via query

client.userinfo(accessToken, { via: 'query' }); // => Promise

auth via body

client.userinfo(accessToken, { verb: 'post', via: 'body' }); // => Promise

userinfo also handles (as long as you have the proper metadata configured) responses that are:

• signed
• signed and encrypted (nested JWT)
• just encrypted

Getting RP-Initiated Logout url

Note: Only usable with issuer's supporting OpenID Connect Session Management 1.0

client.endSessionUrl({
  post_logout_redirect_uri: '...', // OPTIONAL, defaults to client.post_logout_redirect_uris[0] if there's only one
  state: '...', // RECOMMENDED
  id_token_hint: '...', // OPTIONAL, accepts the string value or tokenSet with id_token
}); // => String (URL)

Fetching Distributed Claims

let claims = {
  sub: 'userID',
  _claim_names: {
    credit_history: 'src1',
    email: 'src2',
  },
  _claim_sources: {
    src1: { endpoint: 'https://src1.example.com/claims', access_token: 'foobar' },
    src2: { endpoint: 'https://src2.example.com/claims' },
  },
};

client.fetchDistributedClaims(claims, { src2: 'bearer.for.src2' }) // => Promise
  .then(function (output) {
    console.log('claims %j', claims); // ! also modifies original input, does not create a copy
    console.log('output %j', output);
    // removes fetched names and sources and removes _claim_names and _claim_sources members if they
    // are empty
  });
  // when rejected the error will have a property 'src' with the source name it relates to

Unpacking Aggregated Claims

let claims = {
  sub: 'userID',
  _claim_names: {
    credit_history: 'src1',
    email: 'src2',
  },
  _claim_sources: {
    src1: { JWT: 'probably.a.jwt' },
    src2: { JWT: 'probably.another.jwt' },
  },
};

client.unpackAggregatedClaims(claims) // => Promise, autodiscovers JWT issuers, verifies signatures
  .then(function (output) {
    console.log('claims %j', claims); // ! also modifies original input, does not create a copy
    console.log('output %j', output);
    // removes fetched names and sources and removes _claim_names and _claim_sources members if they
    // are empty
  });
  // when rejected the error will have a property 'src' with the source name it relates to

Custom token endpoint grants

Use when the token endpoint also supports additional grant types.

client.grant({
  grant_type: 'client_credentials',
  scope: 'api:read',
}); // => Promise

client.grant({
  grant_type: 'password',
  username: 'johndoe',
  password: 'A3ddj3w',
  scope: 'profile',
}); // => Promise

Registering new client (via Dynamic Registration)

const opts = { keystore, initialAccessToken }; // both optional
issuer.Client.register(metadata, [opts]) // => opts optional, Promise
  .then(function (client) {
    console.log('Registered client %s, %O', client.client_id, client.metadata);
  });

Generating a signed/encrypted Request Object

client.requestObject({ max_age: 300, redirect_uri })
  .then(function (request) {
    console.log('JWT Request Object %s', request)
  });

This will use the client metadata request_object_signing_alg, request_object_encryption_alg and request_object_encryption_enc, but you can provide the signing and/or encryption algs explicitly

client.requestObject({ max_age: 300, redirect_uri }, {
  // sign: '...',
  // encrypt: {
  //   alg: '...',
  //   enc: '...',
  // }
}).then(function (request) {
  console.log('JWT Request Object %s', request)
});

WebFinger discovery

Issuer.webfinger(userInput) // => Promise
  .then(function (issuer) {
    console.log('Discovered issuer %s %O', issuer.issuer, issuer.metadata);
  });

Accepts, normalizes, discovers and validates the discovery of User Input using E-Mail, URL, acct, Hostname and Port syntaxes as described in Discovery 1.0.

Uses already discovered (cached) issuers where applicable.

TokenSet

authorizationCallback and refresh methods on a Client return TokenSet, when assigned an expires_in value a TokenSet calculates and assigns an expires_at with the corresponding unix time. It also comes with few helpers.

client.authorizationCallback(..., ...).then(function (tokenSet) {
  console.log('tokenSet#expires_at', tokenSet.expires_at);
  console.log('tokenSet#expires_in', tokenSet.expires_in);
  setTimeout(function () {
    console.log('tokenSet#expires_in', tokenSet.expires_in);
  }, 2000);
  console.log('tokenSet#expired()', tokenSet.expired());
  console.log('tokenSet#claims', tokenSet.claims);
});

Usage with passport

Once you have a Client instance, just pass it to the Strategy constructor. Issuer is best discovered, Client passed properties manually or via an uri (see get-started).

Verify function is invoked with a TokenSet, userinfo only when requested, last argument is always the done function which you invoke once you found your user.

const { Strategy } = require('openid-client');
const params = {
  // ... any authorization request parameters go here
  // client_id defaults to client.client_id
  // redirect_uri defaults to client.redirect_uris[0]
  // response type defaults to client.response_types[0], then 'code'
  // scope defaults to 'openid'
}
const passReqToCallback = false; // optional, defaults to false, when true req is passed as a first
                                 // argument to verify fn

const usePKCE = true; // optional, defaults to false, when true the code_challenge_method will be
                      // resolved from the issuer configuration, instead of true you may provide
                      // any of the supported values directly, i.e. "S256" (recommended) or "plain"

passport.use('oidc', new Strategy({ client, [params], [passReqToCallback], [usePKCE] }, (tokenset, userinfo, done) => {
  console.log('tokenset', tokenset);
  console.log('access_token', tokenset.access_token);
  console.log('id_token', tokenset.id_token);
  console.log('claims', tokenset.claims);
  console.log('userinfo', userinfo);

  User.findOne({ id: tokenset.claims.sub }, function (err, user) {
    if (err) return done(err);
    return done(null, user);
  });
}));

// start authentication request
// options [optional], extra authentication parameters
app.get('/auth', passport.authenticate('oidc', [options]));

// authentication callback
app.get('/auth/cb', passport.authenticate('oidc', { successRedirect: '/', failureRedirect: '/login' }));

Configuration

Client Authentication explained

Configure token_endpoint_auth_method with one of the following. Defined in Core 1.0:

• none - only client_id is sent in the request body
• client_secret_basic (default) - client_id and client_secret is sent using the Authorization header as described in RFC6749
• client_secret_post - client_id and client_secret is sent in the request body as described in RFC6749
• client_secret_jwt - using client_secret as a shared symmetrical secret a client_assertion is sent in the request body
• private_key_jwt - using the asymmetric keys provided via keystore a client_assertion is sent in the request body

The configuration may differ between token, introspection and revocation endpoints. The metadata would be:

• token_endpoint_auth_method
• introspection_endpoint_auth_method
• revocation_endpoint_auth_method

The other metadata names follow the same prefix convention.

Note: *_jwt methods resolve their algorithm either via the client's configured alg (token_endpoint_auth_signing_alg) or any of the issuer's supported algs (token_endpoint_auth_signing_alg_values_supported)

Allow for system clock skew

It is possible the RP or OP environment has a system clock skew, to set a clock tolerance (in seconds)

client.CLOCK_TOLERANCE = 5; // to allow a 5 second skew

Changing HTTP request defaults

Setting defaultHttpOptions on Issuer always merges your passed options with the default. openid-client uses got for http requests with the following default request options

const DEFAULT_HTTP_OPTIONS = {
  followRedirect: false,
  headers: { 'User-Agent': `${pkg.name}/${pkg.version} (${pkg.homepage})` },
  retries: 0,
  timeout: 1500,
};

You can add your own headers, change the user-agent used or change the timeout setting

Issuer.defaultHttpOptions = { timeout: 2500, headers: { 'X-Your-Header': '<whatever>' } };

Confirm your httpOptions by

console.log('httpOptions %j', Issuer.defaultHttpOptions);

Proxy settings

Because of the lightweight nature of got library the client will not use environment-defined http(s) proxies. In order to have them used you'll need to either provide your own http request implementation using the provided httpClient setter or use the bundled request one.

Custom implementation:

/*
 * url {String}
 * options {Object}
 * options.headers {Object}
 * options.body {String|Object}
 * options.form {Boolean}
 * options.query {Object}
 * options.timeout {Number}
 * options.retries {Number}
 * options.followRedirect {Boolean}
 */

Issuer.httpClient = {
   get(url, options) {}, // return Promise
   post(url, options) {}, // return Promise
   HTTPError, // used error constructor
};

Bundled (and maintained + tested) request implementation after you've added request to your package.json bundle:

npm install request@^2.0.0 --save

Issuer.useRequest();