AWS JWT Verify

NodeJS library for verifying JWTs signed by Amazon Cognito, and any OIDC-compatible IDP that signs JWTs with RS256.

Installation

npm install aws-jwt-verify

Philosophy of this library

• Do one thing and do it well. Focus solely on verifying JWTs.
• Pure TypeScript library that can be used in NodeJS v12 and above (both CommonJS and ESM supported).
• Support both Amazon Cognito as well as any other OIDC-compatible IDP as first class citizen.
• 0 runtime dependencies, batteries included. This library includes all necessary code to validate RS256-signed JWTs. E.g. it contains a simple (and pluggable) HTTP helper to fetch the JWKS from the JWKS URI, and it includes a simple ASN.1 encoder to transform JWKs into DER-encoded RSA public keys (in order to verify JWTs with NodeJS native crypto calls).
• Opinionated towards the best practices as described by the IETF in JSON Web Token Best Current Practices.
• Make it easy for users to use this library in a secure way. For example, this library requires users to specify issuer and audience, as these should be checked for (see best practices linked to above).

Currently, only signature algorithm RS256 is supported.

Intended Usage

This library was specifically designed to be easy to use in:

• API Gateway Lambda authorizers
• AppSync Lambda authorizers
• CloudFront Lambda@Edge
• NodeJS APIs, e.g. running in AWS Fargate, that need to verify incoming JWTs

Table of Contents

• Verifying JWTs from Amazon Cognito
• Verifying JWTs from any OIDC-compatible IDP
• The JWKS cache
• Usage examples
  • CloudFront Lambda@Edge
  • API Gateway Lambda Authorizer - REST
  • HTTP API Authorizer
  • AppSync Lambda Authorizer
  • Fastify
  • Express
• Security
• License

Verifying JWTs from Amazon Cognito

Create a CognitoJwtVerifier instance and use it to verify JWTs:

1import { CognitoJwtVerifier } from "aws-jwt-verify";
2
3// Verifier that expects valid id tokens:
4const idTokenVerifier = CognitoJwtVerifier.create({
5  userPoolId: "<user_pool_id>",
6  tokenUse: "id",
7  clientId: "<client_id>",
8});
9
10try {
11  const idTokenPayload = await idTokenVerifier.verify(
12    "eyJraWQeyJhdF9oYXNoIjoidk..." // the JWT as string
13  );
14  console.log("Token is valid. Payload:", idTokenPayload);
15} catch {
16  console.log("Token not valid!");
17}

You can also use verifySync, if you've made sure the JWK has already been cached, see further below.

Verification parameters

Except the User Pool ID, parameters provided when creating the CognitoJwtVerifier act as defaults, that can be overridden upon calling verify or verifySync.

Supported parameters are:

• tokenUse (mandatory): verify that the JWT's token_use claim matches your expectation. Set to either id or access. Set to null to skip checking token_use.
• clientId (mandatory): verify that the JWT's aud (id token) or client_id (access token) claim matches your expectation. Provide a string, or an array of strings to allow multiple client ids (i.e. one of these client ids must match the JWT). Set to null to skip checking client id (not recommended unless you know what you are doing).
• group (optional): verify that the JWT's cognito:groups claim matches your expectation. Provide a string, or an array of strings to allow multiple groups (i.e. one of these groups must match the JWT).
• scope (optional): verify that the JWT's scope claim matches your expectation (only of use for access tokens). Provide a string, or an array of strings to allow multiple scopes (i.e. one of these scopes must match the JWT). See also Checking scope.
• graceSeconds (optional, default 0): to account for clock differences between systems, provide the number of seconds beyond JWT expiry (exp claim) or before "not before" (nbf claim) you will allow.
• customJwtCheck (optional): your custom function with additional JWT (and JWK) checks to execute (see also below).

1import { CognitoJwtVerifier } from "aws-jwt-verify";
2
3const verifier = CognitoJwtVerifier.create({
4  userPoolId: "<user_pool_id>", // mandatory, can't be overridden upon calling verify
5  tokenUse: "id", // needs to be specified here or upon calling verify
6  clientId: "<client_id>", // needs to be specified here or upon calling verify
7  group: "admins", // optional
8  graceSeconds: 0, // optional
9  scope: "my-api/read", // optional
10  customJwtCheck: (payload, header, jwk) => {}, // optional
11});
12
13try {
14  const payload = await verifier.verify("eyJraWQeyJhdF9oYXNoIjoidk...", {
15    group: "users", // Cognito group overridden: should be users (not admins)
16  });
17  console.log("Token is valid. Payload:", payload);
18} catch {
19  console.log("Token not valid!");
20}

Checking scope

If you provide scopes to the CognitoJwtVerifier, the verifier will make sure the scope claim in the JWT includes at least one of those scopes:

1import { CognitoJwtVerifier } from "aws-jwt-verify";
2
3const verifier = CognitoJwtVerifier.create({
4  userPoolId: "<user_pool_id>",
5  tokenUse: "access", // scopes are only present on Cognito access tokens
6  clientId: "<client_id>",
7  scope: ["my-api:write", "my-api:admin"],
8});
9
10try {
11  const payload = await verifier.verify("eyJraWQeyJhdF9oYXNoIjoidk...");
12  console.log("Token is valid. Payload:", payload);
13} catch {
14  console.log("Token not valid!");
15}

So a JWT payload like the following would have a valid scope:

1{
2  "client_id": "<client_id>",
3  "scope": "my-api:write someotherscope yetanotherscope", // scope string is split on spaces to gather the array of scopes to compare with
4  "iat": 1234567890,
5  "...": "..."
6}

This scope would not be valid:

1{
2  "client_id": "<client_id>",
3  "scope": "my-api:read someotherscope yetanotherscope", // Neither "my-api:write" nor "my-api:admin" present
4  "iat": 1234567890,
5  "...": "..."
6}

Custom JWT and JWK checks

It's possible to provide a function with your own custom JWT checks. This function will be called if the JWT is valid, at the end of the JWT verification.

The function will be called with:

• the decoded JWT header
• the decoded JWT payload
• the JWK that was used to verify the JWT

Throw an error in this function if you want to reject the JWT.

1import { CognitoJwtVerifier } from "aws-jwt-verify";
2
3const idTokenVerifier = CognitoJwtVerifier.create({
4  userPoolId: "<user_pool_id>",
5  tokenUse: "id",
6  clientId: "<client_id>",
7  customJwtCheck: async ({ header, payload, jwk }) => {
8    if (header.someHeaderField !== "expected") {
9      throw new Error("something wrong with the header");
10    }
11    if (payload.somePayloadField !== "expected") {
12      throw new Error("something wrong with the payload");
13    }
14    if (jwk.someJwkfField !== "expected") {
15      throw new Error("something wrong with the jwk");
16    }
17    await someAsyncCheck(...); // can call out to a DB or do whatever
18  },
19});
20
21// This will now throw, even if the JWT is otherwise valid, if your custom function throws:
22await idTokenVerifier.verify("eyJraWQeyJhdF9oYXNoIjoidk...");

Note that customJwtCheck may be an async function, but only if you use verify (not supported for verifySync).

Trusting multiple User Pools

If you want to allow JWTs from multiple User Pools, provide an array with these User Pools upon creating the verifier:

1import { CognitoJwtVerifier } from "aws-jwt-verify";
2
3// This verifier will trust both User Pools
4const idTokenVerifier = CognitoJwtVerifier.create([
5  {
6    userPoolId: "<user_pool_id>",
7    tokenUse: "id",
8    clientId: "<client_id>", // clientId is mandatory at verifier level now, to disambiguate between User Pools
9  },
10  {
11    userPoolId: "<user_pool_id_2>",
12    tokenUse: "id",
13    clientId: "<client_id_2>",
14  },
15]);
16
17try {
18  const idTokenPayload = await idTokenVerifier.verify(
19    "eyJraWQeyJhdF9oYXNoIjoidk..." // token must be signed by either of the User Pools
20  );
21  console.log("Token is valid. Payload:", idTokenPayload);
22} catch {
23  console.log("Token not valid!");
24}

Using the generic JWT RSA verifier for Cognito JWTs

The generic JwtRsaVerifier (see below) can also be used for Cognito, which is useful if you want to define a verifier that trusts multiple IDPs, i.e. Cognito and another IDP:

1import { JwtRsaVerifier } from "aws-jwt-verify";
2
3const verifier = JwtRsaVerifier.create([
4  {
5    issuer: "https://cognito-idp.eu-west-1.amazonaws.com/<user_pool_id>",
6    audience: "<client_id>",
7  },
8  {
9    issuer: "https://example.com/my/other/idp",
10    audience: "myaudience",
11  },
12]);

Verifying JWTs from any OIDC-compatible IDP

The generic JwtRsaVerifier works for any OIDC-compatible IDP that signs JWTs with RS256:

1import { JwtRsaVerifier } from "aws-jwt-verify";
2
3const verifier = JwtRsaVerifier.create({
4  issuer: "https://example.com/", // set this to the expected "iss" claim on your JWTs
5  audience: "<audience>", // set this to the expected "aud" claim on your JWTs
6});
7
8try {
9  const payload = await verifier.verify("eyJraWQeyJhdF9oYXNoIjoidk...");
10  console.log("Token is valid. Payload:", payload);
11} catch {
12  console.log("Token not valid!");
13}

Support Multiple IDP's:

1const verifier = JwtRsaVerifier.create([
2  {
3    issuer: "https://example.com/idp1",
4    audience: "expectedAudienceIdp1",
5  },
6  {
7    issuer: "https://example.com/idp2",
8    audience: "expectedAudienceIdp2",
9  },
10]);
11
12try {
13  const otherPayload = await verifier.verify("eyJraWQeyJhdF9oYXNoIjoidk..."); // Token must be from either idp1 or idp2
14  console.log("Token is valid. Payload:", otherPayload);
15} catch {
16  console.log("Token not valid!");
17}

Supported verify parameters

Except issuer, parameters provided when creating the JwtRsaVerifier act as defaults, that can be overridden upon calling verify or verifySync.

Supported parameters are:

• jwksUri (optional, can only be provided at verifier level): the URI where the JWKS can be downloaded from. To find this URI for your IDP, consult your IDP's OpenId configuration (e.g. by opening the OpenId configuration in your browser). Usually, it is ${issuer}/.well-known/jwks.json, which is the default value that will be used if you don't explicitly provide jwksUri.
• audience (mandatory): verify that the JWT's aud claim matches your expectation. Provide a string, or an array of strings to allow multiple client ids (i.e. one of these audiences must match the JWT). Set to null to skip checking audience (not recommended unless you know what you are doing). Note that a JWT's aud claim might be an array of audiences. The JwtRsaVerifier will in that case make sure that at least one of these audiences matches with at least one of the audiences that were provided to the verifier.
• scope (optional): verify that the JWT's scope claim matches your expectation (only of use for access tokens). Provide a string, or an array of strings to allow multiple scopes (i.e. one of these scopes must match the JWT). See also Checking scope.
• graceSeconds (optional, default 0): to account for clock differences between systems, provide the number of seconds beyond JWT expiry (exp claim) or before "not before" (nbf claim) you will allow.
• customJwtCheck (optional): your custom function with additional JWT checks to execute (see Custom JWT and JWK checks).

The JWKS cache

The JWKS cache is responsible for fetching the JWKS from the JWKS URI, caching it, and selecting the right JWK from it. Both the CognitoJwtVerifier and the (generic) JwtRsaVerifier utilize an in-memory JWKS cache. For each issuer a JWKS cache is maintained, and each JWK in a JWKS is selected and cached using its kid (key id). The JWKS for an issuer will be fetched once initially, and thereafter only upon key rotations (detected by the occurrence of a JWT with a kid that is not yet in the cache).

Note: examples below work the same for CognitoJwtVerifier and JwtRsaVerifier.

Loading the JWKS from file

If e.g. your runtime environment doesn't have internet access, or you want to prevent the fetch over the network, you can load the JWKS explicitly yourself:

1import { CognitoJwtVerifier } from "aws-jwt-verify";
2import { readFileSync } from "fs";
3
4const idTokenVerifier = CognitoJwtVerifier.create({
5  userPoolId: "<user_pool_id>",
6  tokenUse: "id",
7  clientId: "<client_id>",
8});
9
10const jwks = JSON.parse(readFileSync("jwks.json", { encoding: "utf-8" }));
11idTokenVerifier.cacheJwks(jwks);
12
13// Because the JWKS doesn't need to be downloaded now, you can use verifySync:
14try {
15  const idTokenPayload = idTokenVerifier.verifySync(
16    "eyJraWQeyJhdF9oYXNoIjoidk..."
17  );
18  console.log("Token is valid. Payload:", payload);
19} catch {
20  console.log("Token not valid!");
21}
22
23// Async verify will of course work as well (and will use the cache also):
24try {
25  const idTokenPayload = await idTokenVerifier.verify(
26    "eyJraWQeyJhdF9oYXNoIjoidk..."
27  );
28  console.log("Token is valid. Payload:", idTokenPayload);
29} catch {
30  console.log("Token not valid!");
31}

Note that the verifier will still try to fetch the JWKS, if it encounters a JWT with a kid that is not in it's cached JWKS (i.e. to cater for key rotations).

Rate limiting

Both the CognitoJwtVerifier and the JwtRsaVerifier enforce a rate limit of 1 JWKS download per JWKS uri per 10 seconds. This protects users of this library from inadvertently flooding the JWKS uri with requests, and prevents wasting time doing network calls.

The rate limit works as follows (implemented by the penaltyBox, see below). When the verifier fetches the JWKS and fails to locate the JWT's kid in the JWKS, an error is thrown, and a timer of 10 seconds is started. Until that timer completes, the verifier will refuse to fetch the particular JWKS uri again. It will instead throw an error immediately on verify calls where that would require the JWKS to be downloaded.

The verifier will continue to verify JWTs for which the right JWK is already present in the cache, also it will still try other JWKS uris (for other issuers).

It is possible to implement a different rate limiting scheme yourself, by customizing the JWKS cache, or the penaltyBox implementation, see below.

Explicitly hydrating the JWKS cache

In a long running NodeJS API (e.g. a Fargate container), it might make sense to hydrate the JWKS cache upon server start up. This will speed up the first JWT verification, as the JWKS doesn't have to be downloaded anymore.

This call will always fetch the current, latest, JWKS for each of the verifier's issuers (even though the JWKS might have been fetched and cached before):

1const verifier = JwtRsaVerifier.create([
2  {
3    issuer: "https://example.com/idp1",
4    audience: "myappclient1",
5  },
6  {
7    issuer: "https://example.com/idp2",
8    audience: "myappclient2",
9  },
10]);
11
12// Fetch and cache the JWKS for all configured issuers
13await verifier.hydrate();

Note: it is only useful to call this method if your calling process has an idle time window, in which it might just as well fetch the JWKS. For example, during container start up, when the load balancer does not yet route traffic to the container. Calling this method inside API Gateway custom authorizers or Lambda@Edge has no benefit (in fact, awaiting the call as part of the Lambda handler would even hurt performance as it bypasses the existing cached JWKS).

Clearing the JWKS cache

If you have a predefined rotation schedule for your JWKS, you could set the refresh interval of the verifier aligned to this schedule:

1import { JwtRsaVerifier } from "aws-jwt-verify";
2
3const verifier = JwtRsaVerifier.create({
4  issuer: "https://example.com/",
5  audience: "<audience>",
6});
7
8setInterval(() => {
9  verifier.cacheJwks({ keys: [] }); // empty cache, by loading an empty JWKS
10}, 1000 * 60 * 60 * 4); // For a 4 hour refresh schedule

If an automated rotation does not fit your use case, and you need to clear out the JWKS cache, you could use:

1verifier.cacheJwks({ keys: [] });

Customizing the JWKS cache

When you instantiate a CognitoJwtVerifier or JwtRsaVerifier without providing a JwksCache, the SimpleJwksCache is used:

1import { JwtRsaVerifier } from "aws-jwt-verify";
2import { SimpleJwksCache } from "aws-jwt-verify/jwk";
3
4const verifier = JwtRsaVerifier.create({
5  issuer: "http://my-tenant.my-idp.com",
6});
7
8// Equivalent:
9const verifier2 = JwtRsaVerifier.create(
10  {
11    issuer: "http://my-tenant.my-idp.com",
12  },
13  {
14    jwksCache: new SimpleJwksCache(),
15  }
16);

The SimpleJwksCache can be tailored by using a different penaltyBox and/or fetcher (see below).

Alternatively, you can implement an entirely custom JwksCache yourself, by creating a class that implements the interface JwksCache (from "aws-jwt-verify/jwk"). This allows for highly custom scenario's, e.g. you could implement a JwksCache with custom logic for selecting a JWK from the JWKS.

Sharing the JWKS cache amongst different verifiers

If you want to define multiple verifiers for the same JWKS uri, it makes sense to share the JWKS cache, so the JWKS will be downloaded and cached once:

1import { JwtRsaVerifier } from "aws-jwt-verify";
2import { SimpleJwksCache } from "aws-jwt-verify/jwk";
3
4const sharedJwksCache = new SimpleJwksCache();
5
6const verifierA = JwtRsaVerifier.create(
7  {
8    jwksUri: "https://example.com/keys/jwks.json",
9    issuer: "https://example.com/",
10    audience: "<audience>",
11  },
12  {
13    jwksCache: sharedJwksCache,
14  }
15);
16
17const verifierB = JwtRsaVerifier.create(
18  {
19    jwksUri: "https://example.com/keys/jwks.json", // same JWKS URI, so sharing cache makes sense
20    issuer: "https://example.com/",
21    audience: "<audience>",
22  },
23  {
24    jwksCache: sharedJwksCache,
25  }
26);

Using a different JsonFetcher with SimpleJwksCache

When instantiating SimpleJwksCache, the fetcher property can be populated with an instance of a class that implements the interface JsonFetcher (from "aws-jwt-verify/https"), such as the SimpleJsonFetcher (which is the default).

The purpose of the fetcher, is to execute fetches against the JWKS uri (HTTPS GET) and parse the resulting JSON file. The default implementation, the SimpleJsonFetcher, has basic machinery to do fetches over HTTPS. It does 1 (immediate) retry in case of connection errors.

By supplying a custom fetcher when instantiating SimpleJwksCache, instead of SimpleJsonFetcher, you can implement any retry and backoff scheme you want, or use another HTTPS library:

1import { JwtRsaVerifier } from "aws-jwt-verify";
2import { SimpleJwksCache } from "aws-jwt-verify/jwk";
3import { JsonFetcher } from "aws-jwt-verify/https";
4import axios from "axios";
5
6// Use axios to do the HTTPS fetches
7class CustomFetcher implements JsonFetcher {
8  instance = axios.create();
9  public async fetch(uri: string) {
10    return this.instance.get(uri).then((response) => response.data);
11  }
12}
13
14const verifier = JwtRsaVerifier.create(
15  {
16    issuer: "http://my-tenant.my-idp.com",
17  },
18  {
19    jwksCache: new SimpleJwksCache({
20      fetcher: new CustomFetcher(),
21    }),
22  }
23);

Using a different penaltyBox with SimpleJwksCache

When instantiating SimpleJwksCache, the penaltyBox property can be populated with an instance of a class that implements the interface PenaltyBox (from "aws-jwt-verify/jwk"), such as the SimplePenaltyBox (which is the default).

The SimpleJwksCache will always do await penaltyBox.wait(jwksUri, kid) before asking the fetcher to fetch the JWKS.

By supplying a custom penaltyBox when instantiating SimpleJwksCache, instead of SimplePenaltyBox, you can implement any waiting scheme you want, in your implementation of the wait function.

The SimpleJwksCache will call penaltyBox.registerSuccessfulAttempt(jwksUri, kid) when it succeeds in locating the right JWK in the JWKS, and call penaltyBox.registerFailedAttempt(jwksUri, kid) otherwise. You need to process these calls, so that you can determine the right amount of waiting in your wait implementation.

1import { JwtRsaVerifier } from "aws-jwt-verify";
2import {
3  SimpleJwksCache,
4  SimplePenaltyBox,
5  PenaltyBox,
6} from "aws-jwt-verify/jwk";
7
8// In this example we use the SimplePenaltyBox, but override the default wait period
9const verifier = JwtRsaVerifier.create(
10  {
11    issuer: "http://my-tenant.my-idp.com",
12  },
13  {
14    jwksCache: new SimpleJwksCache({
15      penaltyBox: new SimplePenaltyBox({ waitSeconds: 1 }),
16    }),
17  }
18);
19
20// Or implement your own penaltyBox
21// The example here just stupidly waits 5 second always,
22// even on the first fetch of the JWKS uri
23class CustomPenaltyBox implements PenaltyBox {
24  public async wait(jwksUri: string, kid: string) {
25    // implement something better
26    await new Promise((resolve) => setTimeout(resolve, 5000));
27  }
28  public release(jwksUri: string, kid?: string) {
29    // implement
30  }
31  public registerFailedAttempt(jwksUri: string, kid: string) {
32    // implement
33  }
34  public registerSuccessfulAttempt(jwksUri: string, kid: string) {
35    // implement
36  }
37}
38const verifier2 = JwtRsaVerifier.create(
39  {
40    issuer: "http://my-tenant.my-idp.com",
41  },
42  {
43    jwksCache: new SimpleJwksCache({ penaltyBox: new CustomPenaltyBox() }),
44  }
45);

Usage Examples

CloudFront Lambda@Edge

The verifier should be instantiated outside the Lambda handler, so the verifier's cache can be reused for subsequent requests for as long as the Lambda functions stays "hot".

This is an example of a Viewer Request Lambda@Edge function, that inspects each incoming request. It requires each incoming request to have a valid JWT (in this case an access token that includes scope "read") in the HTTP "Authorization" header.

1const { CognitoJwtVerifier } = require("aws-jwt-verify");
2
3// Create the verifier outside the Lambda handler (= during cold start),
4// so the cache can be reused for subsequent invocations. Then, only during the
5// first invocation, will the verifier actually need to fetch the JWKS.
6const jwtVerifier = CognitoJwtVerifier.create({
7  userPoolId: "<user_pool_id>",
8  tokenUse: "access",
9  clientId: "<client_id>",
10  scope: "read",
11});
12
13exports.handler = async (event) => {
14  const accessToken = event.Records[0].cf.request.headers["authorization"];
15  try {
16    await jwtVerifier.verify(accessToken);
17  } catch {
18    return {
19      status: "403",
20      body: "Unauthorized",
21    };
22  }
23  return request; // allow request to proceed
24};

API Gateway Lambda Authorizer - REST

The verifier should be instantiated outside the Lambda handler, so the verifier's cache can be reused for subsequent requests for as long as the Lambda functions stays "hot".

Two types of API Gateway Lambda authorizers could be created - token based and request-based. For both the types of authorizers, you could use the AWS API Gateway Lambda Authorizer BluePrint as a reference pattern where the token validation could be achieved as follows

For token based authorizers, where lambda event payload is set to Token and token source is set to (http) Header with name authorization:

1const { CognitoJwtVerifier } = require("aws-jwt-verify");
2
3// Create the verifier outside the Lambda handler (= during cold start),
4// so the cache can be reused for subsequent invocations. Then, only during the
5// first invocation, will the verifier actually need to fetch the JWKS.
6const jwtVerifier = CognitoJwtVerifier.create({
7  userPoolId: "<user_pool_id>",
8  tokenUse: "access",
9  clientId: "<client_id>",
10  scope: "read",
11});
12
13exports.handler = async (event) => {
14  const accessToken = event.authorizationToken;
15
16  let payload;
17  try {
18    // If the token is not valid, an error is thrown:
19    payload = await jwtVerifier.verify(accessToken);
20  } catch {
21    // API Gateway wants this *exact* error message, otherwise it returns 500 instead of 401:
22    throw new Error("Unauthorized");
23  }
24
25  // Proceed with additional authorization logic
26  // ...
27};

For request based authorizers, where lambda event payload is set to Request and identity source is set to (http) Header with name authorization:

1const { CognitoJwtVerifier } = require("aws-jwt-verify");
2
3// Create the verifier outside the Lambda handler (= during cold start),
4// so the cache can be reused for subsequent invocations. Then, only during the
5// first invocation, will the verifier actually need to fetch the JWKS.
6const jwtVerifier = CognitoJwtVerifier.create({
7  userPoolId: "<user_pool_id>",
8  tokenUse: "access",
9  clientId: "<client_id>",
10  scope: "read",
11});
12
13exports.handler = async (event) => {
14  const accessToken = event.headers["authorization"];
15
16  let payload;
17  try {
18    // If the token is not valid, an error is thrown:
19    payload = await jwtVerifier.verify(accessToken);
20  } catch {
21    // API Gateway wants this *exact* error message, otherwise it returns 500 instead of 401:
22    throw new Error("Unauthorized");
23  }
24
25  // Proceed with additional authorization logic
26  // ...
27};

HTTP API Lambda Authorizer

An example of a sample HTTP Lambda authorizer is included here as part of the test suite for the solution (format 2.0).

AppSync Lambda Authorizer

The verifier should be instantiated outside the Lambda handler, so the verifier's cache can be reused for subsequent requests for as long as the Lambda functions stays "hot".

This is an example of AppSync Lambda Authorization function, that validates the JWT is valid (in this case an access token that includes scope "read") along with other authorization business logic

1const { CognitoJwtVerifier } = require("aws-jwt-verify");
2
3// Create the verifier outside the Lambda handler (= during cold start),
4// so the cache can be reused for subsequent invocations. Then, only during the
5// first invocation, will the verifier actually need to fetch the JWKS.
6const jwtVerifier = CognitoJwtVerifier.create({
7  userPoolId: "<user_pool_id>",
8  tokenUse: "access",
9  clientId: "<client_id>",
10  scope: "read",
11});
12
13exports.handler = async (event) => {
14  const accessToken = event.authorizationToken;
15  try {
16    await jwtVerifier.verify(accessToken);
17  } catch {
18    return {
19      isAuthorized: False,
20    };
21  }
22  //Proceed with additional authorization logic
23};

Fastify

1const { CognitoJwtVerifier } = require("aws-jwt-verify");
2const fastify = require("fastify")({ logger: true });
3
4// Create the verifier outside your route handlers,
5// so the cache is persisted and can be shared amongst them.
6const jwtVerifier = CognitoJwtVerifier.create({
7  userPoolId: "<user_pool_id>",
8  tokenUse: "access",
9  clientId: "<client_id>",
10  scope: "read",
11});
12
13fastify.get("/", async (request, reply) => {
14  try {
15    // A valid JWT is expected in the HTTP header "authorization"
16    await jwtVerifier.verify(request.headers.authorization);
17  } catch (authErr) {
18    fastify.log.error(authErr);
19    const err = new Error();
20    err.statusCode = 403;
21    throw err;
22  }
23  return { private: "only visible to users sending a valid JWT" };
24});
25
26const startFastify = async () => {
27  try {
28    await fastify.listen(3000);
29  } catch (err) {
30    fastify.log.error(err);
31    process.exit(1);
32  }
33};
34
35// Hydrate the JWT verifier, and start Fastify.
36// Hydrating the verifier makes sure the JWKS is loaded into the JWT verifier,
37// so it can verify JWTs immediately without any latency.
38// (Alternatively, just start Fastify, the JWKS will be downloaded when the first JWT is being verified then)
39Promise.all([jwtVerifier.hydrate(), () => fastify.listen(3000)]).catch(
40  (err) => {
41    fastify.log.error(err);
42    process.exit(1);
43  }
44);

Express

1const { CognitoJwtVerifier } = require("aws-jwt-verify");
2const express = require("express");
3const app = express();
4const port = 3000;
5
6// Create the verifier outside your route handlers,
7// so the cache is persisted and can be shared amongst them.
8const jwtVerifier = CognitoJwtVerifier.create({
9  userPoolId: "<user_pool_id>",
10  tokenUse: "access",
11  clientId: "<client_id>",
12  scope: "read",
13});
14
15app.get("/", async (req, res, next) => {
16  try {
17    // A valid JWT is expected in the HTTP header "authorization"
18    await jwtVerifier.verify(req.header("authorization"));
19  } catch (err) {
20    console.error(err);
21    return res.status(403).json({ statusCode: 403, message: "Forbidden" });
22  }
23  res.json({ private: "only visible to users sending a valid JWT" });
24});
25
26// Hydrate the JWT verifier, then start express.
27// Hydrating the verifier makes sure the JWKS is loaded into the JWT verifier,
28// so it can verify JWTs immediately without any latency.
29// (Alternatively, just start express, the JWKS will be downloaded when the first JWT is being verified then)
30jwtVerifier
31  .hydrate()
32  .catch((err) => {
33    console.error(`Failed to hydrate JWT verifier: ${err}`);
34    process.exit(1);
35  })
36  .then(() =>
37    app.listen(port, () => {
38      console.log(`Example app listening at http://localhost:${port}`);
39    })
40  );

Security

See CONTRIBUTING for more information.

License

This project is licensed under the Apache-2.0 License.