Setting up JWT Authentication

This article explains how to control authentication of your web resources using JWT authentication.

Introduction

With NGINX Plus it is possible to control access to your resources using JWT authentication. JWT is data format for user information in the OpenID Connect standard, which is the standard identity layer on top of the OAuth 2.0 protocol. Deployers of APIs and microservices are also turning to the JWT standard for its simplicity and flexibility. With JWT authentication, a client provides a JSON Web Token, and the token will be validated against a local key file or a remote service.

Prerequisites

NGINX Plus supports the following types of JWT:

  • JSON Web Signature (JWS) - JWT content is digitally signed. The following algorithms can be used for signing:

    • HS256, HS384, HS512
    • RS256, RS384, RS512
    • ES256, ES384, ES512
    • EdDSA (Ed25519 and Ed448 signatures)
  • JSON Web Encryption (JWE) - the contents of JWT is encrypted. The following content encryption algorithms (the “enc” field of JWE header) are supported:

    • A128CBC-HS256, A192CBC-HS384, A256CBC-HS512
    • A128GCM, A192GCM, A256GCM

    The following key management algorithms (the “alg” field of JWE header) are supported:

    • A128KW, A192KW, A256KW
    • A128GCMKW, A192GCMKW, A256GCMKW
    • dir - direct use of a shared symmetric key as the content encryption key
    • RSA-OAEP, RSA-OAEP-256, RSA-OAEP-384, RSA-OAEP-512
  • Nested JWT - support for JWS enclosed into JWE

Configuring NGINX Plus to Authenticate API

Let’s assume that NGINX Plus serves as a gateway (proxy_pass http://api_server) to a number of API servers (the upstream {} block), and requests passed to the API servers should be authenticated:

upstream api_server {
    server 10.0.0.1;
    server 10.0.0.2;
}

server {
    listen 80;

    location /products/ {
        proxy_pass http://api_server;
        #...
    }
}

To implement JWT for authentication:

  1. First, it is necessary to create a JWT that will be issued to a client. You can use your identity provider (IdP) or your own service to create JWTs. For testing purposes, you can create your own JWT, see Authenticating API Clients with JWT and NGINX Plus blog post for details.

  2. Configure NGINX Plus to accept JWT: specify the auth_jwt directive that enables JWT authentication and also defines the authentication area (or “realm”, “API” in the example):

    server {
        listen 80;
    
        location /products/ {
            proxy_pass http://api_server;
            auth_jwt   "API";
            #...
        }
    }
    

    NGINX Plus can also obtain the JWT from a query string parameter. To configure this, include the token= parameter to the auth_jwt directive:

       #...
       auth_jwt "API" token=$arg_apijwt;
       #...
    
  3. Specify the type of JWT - signed (JWS), encrypted (JWE) or nested (Nested JWT) - with the auth_jwt_type directive. The default value of the directive is signed, so for JWS, the directive can be omitted.

    server {
        listen 80;
    
        location /products/ {
            proxy_pass        http://api_server;
            auth_jwt          "API";
            auth_jwt_type     encrypted;
            #...
        }
    }
    
  4. Specify the path to the JSON Web Key file that will be used to verify JWT signature or decrypt JWT content, depending on what you are using. This can be done with the auth_jwt_key_file and/or auth_jwt_key_request directives. Specifying both directives at the same time will allow you to specify more than one source for keys. If none of the directives are specified, JWS signature verification will be skipped.

    In this scenario, the keys will be taken from two files: the key.jwk file and the keys.json file:

    server {
        listen 80;
    
        location /products/ {
            proxy_pass        http://api_server;
            auth_jwt          "API";
            auth_jwt_type     encrypted;
            auth_jwt_key_file conf/key.jwk;
            auth_jwt_key_file conf/keys.json;
        }
    }
    

    In this scenario, there are also two sources for the keys, but the private keys will be taken from the local file private_jwe_keys.jwk, while the public keys will be taken from the external identity provider service https://idp.example.com in a subrequest:

    server {
        listen 80;
    
        location /products/ {
            proxy_pass           http://api_server;
            auth_jwt             "API";
            auth_jwt_type        encrypted;
            auth_jwt_key_file    private_jwe_keys.jwk;
            auth_jwt_key_request /public_jws_keys;
        }
    
        location /public_jws_keys {
            proxy_pass "https_//idp.example.com/keys";
        }
    }
    

    It is recommended to enable JWT key caching to get the optimal performance from the JWT module. For example, you can use the auth_jwt_key_cache directive for the above configuration, and enable the JWT key caching for one hour. Note that if the auth_jwt_key_request or auth_jwt_key_file are configured dynamically with variables, auth_jwt_key_cache cannot be used.

    server {
        listen 80;
    
        location /products/ {
            proxy_pass           http://api_server;
            auth_jwt             "API";
            auth_jwt_type        encrypted;
            auth_jwt_key_file    private_jwe_keys.jwk;
            auth_jwt_key_request /public_jws_keys;
            auth_jwt_key_cache   1h;
        }
    
        location /public_jws_keys {
            proxy_pass "https_//idp.example.com/keys";
        }
    }
    

How NGINX Plus Validates a JWT

A JWT is considered to be valid when the following conditions are met:

  • The signature can be verified (for JWS) or payload can be decrypted (for JWE) with the key found in the auth_jwt_key_file or auth_jwt_key_request (matching on the kid (“key ID”), if present, and alg (“algorithm”) header fields).
  • The JWT is presented inside the validity period, when defined by one or both of the nbf (“not before”) and exp (“expires”) claims.

Creating a JSON Web Key File

In order to validate the signature with a key or to decrypt data, a JSON Web Key (key.jwk) should be created. The file format is defined by JSON Web Key specification:

{"keys":
    [{
        "k":"ZmFudGFzdGljand0",
        "kty":"oct",
        "kid":"0001"
    }]
}

where:

  • the k field is the generated symmetric key (base64url-encoded) basing on a secret (fantasticjwt in the example). The secret can be generated with the following command:
$ echo -n fantasticjwt | base64 | tr '+/' '-_' | tr -d '='
ZmFudGFzdGljand0
  • the kty field defines the key type as a symmetric key (octet sequence)
  • the kid (Key ID) field defines a serial number for this JSON Web Key

Getting JWKs from Subrequest

NGINX Plus can be configured to fetch JSON Web Keys from the remote location - usually an identity provider, especially when using OpenID Connect. The IdP URI where the subrequest will be sent to is configured with the auth_jwt_key_request directive:

http {
    #...

    server {
        listen 80;
            #...

        location / {
            auth_jwt "closed site";
            auth_jwt_key_request /_jwks_uri; # Keys will be fetched by subrequest

            proxy_pass http://my_backend;
        }
    }
}

The URI may refer to an internal location (_jwks_uri) so that the JSON Web Key Set can be cached (proxy_cache and proxy_cache_path directives) to avoid validation overhead. Turning on caching is recommended for high-load API gateways even if JWT key caching is used as it will help to avoid overwhelming a key server with key requests when a JWT key cache expires.

http {
    proxy_cache_path /var/cache/nginx/jwk levels=1 keys_zone=jwk:1m max_size=10m;
    #...

    server {
        listen 80;
            #...

        location = /_jwks_uri {
            internal;
            proxy_method      GET;
            proxy_cache       jwk; # Cache responses
            proxy_cache_valid 200 12h;
            proxy_pass        https://idp.example.com/oauth2/keys; # Obtain keys from here
        }
    }
}

The full example of getting JWKs from a subrequest:

#
proxy_cache_path /var/cache/nginx/jwk levels=1 keys_zone=jwk:1m max_size=10m;

server {
    listen 80; # Use SSL/TLS in production

    location / {
        auth_jwt             "closed site";
        auth_jwt_key_cache   1h;
        auth_jwt_key_request /_jwks_uri;    # Keys will be fetched by subrequest

        proxy_pass http://my_backend;
    }

    location = /_jwks_uri {
        internal;
        proxy_method      GET;
        proxy_cache       jwk; # Cache responses
        proxy_cache_valid 200 12h;
        proxy_pass        https://idp.example.com/oauth2/keys; # Obtain keys from here
    }
}

Arbitrary JWT Claims Validation

During JWT verification, NGINX Plus automatically validates only nbf (“not before”) and exp (“expires”) claims. However, in some cases you need to set more conditions for a successful JWT validation, in particular when dealing with application-specific or protocol level claims. For example, OpenID Connect Core requires validation of iss (“issuer”), aud (“audience”), sub (“subject”) claims for ID token.

Additional conditions for JWT validation can be set as variables with the map module and then evaluated with the auth_jwt_require directive.

In this scenario, we are verifying that:

  • the recipient of the token (audience) is our APIs (map rule 1)
  • the token was issued by a trusted identity provider (map rule 2)
  • scopes in APIs called on behalf of administrators (map rule 3)

The values of three resulting variables are evaluated in the auth_jwt_require directive, and if the value of each variable is 1, the JWT will be accepted:

upstream api_server {
    server 10.0.0.1;
    server 10.0.0.2;
}

map $jwt_claim_aud $valid_app_id {    #map rule 1:
    "~api\d.example.com" 1;           #token issued only for target apps
}

map $jwt_claim_iss $valid_issuer {    #map rule 2:
    "https://idp.example.com/sts" 1;  #token issued by trusted CA
}

map $jwt_claim_scope $valid_scope {   #map rule 3:
    "access_as_admin" 1;              #access as admin only
}

server {
    listen 80;

    location /products/ {
        auth_jwt          "API";
        auth_jwt_key_file conf/api_secret.jwk;
        auth_jwt_require  $valid_app_id $valid_issuer $valid_scope;
        proxy_pass        http://api_server;
    }
}

In some cases the auth_jwt_require directive can be specified multiple times, for example, for the purpose of authentication and then for authorization. In case of an error, the 401 code will be displayed. Assigning the custom error code 403 to another auth_jwt_require directive makes ti possible to differentiate authentication and authorization usecases and handle corresponding failures appropriately:

    location /products/ {
        auth_jwt          "API";
        auth_jwt_key_file conf/api_secret.jwk;
        auth_jwt_require  $valid_app_id $valid_issuer $valid_scope;
        auth_jwt_require  $valid_scope error=403;
        proxy_pass        http://api_server;
    }

Nested JWT Extraction

A Nested JWT is a JWS token enclosed into JWE. In a Nested JWT, the sensitive information from JWS is protected with extra encryption of JWE.

Using Nested JWT may be preferable over JWE because:

  • in case of JWE, the target application/service needs to decrypt the token first, then verify the signature. Decrypt operation on the application side may be time and resource consuming.

  • in case of Nested JWT, as NGINX Plus resides in the same trusted network with the target application, there is no need for token encryption between NGINX Plus and the application. NGINX Plus decrypts the JWE, checks the enclosed JWS, and sends the Bearer Token to the application. This will offload JWE decryption from the application to NGINX Plus.

  • if your application doesn’t support JWE, using Nested JWT enables full protection for JWS.

To enable Nested tokens:

  1. Specify the nested type of JWT with the auth_jwt_type directive.
auth_jwt_type nested;
  1. Pass the decrypted payload (the $jwt_payload variable) to the application as the Bearer token value in the Authorization header:
proxy_set_header Authorization "Bearer $jwt_payload";

This example sums up the previous steps into one configuration:

upstream api_server {
    server 10.0.0.1;
    server 10.0.0.2;
}

http {
    server {
        listen 80;

        auth_jwt          "API";
        auth_jwt_type     nested;
        auth_jwt_key_file conf/api_secret.jwk;

        proxy_pass       http://api_server;
        proxy_set_header Authorization "Bearer $jwt_payload";
    }
}

See Also