When developing cloud-native microservices, we need to think about securing the data that is being propagated from one service to another service and securing the data at rest. So how much security is enough to secure our architecture? Is it the user that identifies itself and decides what data he has access to?


Our cloud-native architecture

In this blog series we will cover these questions and guide you in applying the security layer to your cloud-native blueprint. With this blueprint, we are going to use the Spring ecosystem throughout the series. Solving the following problems is crucial for building a cloud-native microservices architecture, but it should be technology-agnostic:


Where our journey begins…

When it comes to users interacting with our system, we want to verify that the person can identify him- or herself. Most of the time this appears in a login form where you enter your credentials, or in a login page from a third party application (Facebook, Google, etc).


If you like more secure systems, you can add another level of complexity on top of it. Most commonly used is Two-factor-authentication, where the client will use an external provider (Google Authenticator for example) to issue a token for your registered application.


Authorization is the mechanism that uses the user’s data to verify what he is allowed to do.

For instance, who has access to which resources and what are his access rights (eg. read or write) to those resources.

To use these two mechanisms in our system, we will be using a security protocol that fits our microservices architecture. Since we don’t want everyone to have an account for each (micro)service, we aim to have one single identity per person so that the user needs to authenticate only once.

Using the OAuth2 Protocol

When searching for a security protocol, we don’t want to reinvent the wheel and look at what is supported by the Spring framework. Obviously, it depends on the use case of the applications that require resources from our system. Is it a third party application like Facebook or a first party like your own application? Or both? I will explain both OAuth2 and JSON Web Token and how they solve these requirements.

The OAuth2 delegation protocol allows us to retrieve an access token from an identity provider and gain access to a microservice by passing the token with subsequent requests. When introducing the OAuth2 framework to our system, we will be using four grant types. These grant types are different ways to obtain an access token, some clients are more trusted than others.

OAuth2 Grant Types

Third party applications: Authorization Code grant type and Implicit grant type

Authorization Code is the most common used grant type for third party applications, where user’s confidentiality can be maintained. The user won’t have to share his credentials with the application that is requesting resources from our backend. This is a redirection-based flow, which means that the application must be capable of interacting with the user’s web browser.

  • The frontend (application) makes a request to the User Authentication & Authorization server (UAA) on behalf of the user
  • The UAA server redirects to a permission window of a third party for the user to grant permission, the user authenticates and grants permission
  • The UAA server returns an authorization code with a redirect url
  • The frontend uses the authorization code and an application identification to request an access token from the UAA server
  • The UAA verifies the authorization code and returns an access token

Implicit grant type follows the same principle as the Authorization Code type but does not exchange an authorization code to issue an access token.

First party applications: Password grant type

This grant type is best used for first party applications, where the user is in a trust relationship with the application. The application authenticates on behalf of the user and receives the proper JWT.

  • The user provides his credentials to the frontend, commonly done with a login form
  • The frontend assembles a POST request with the credentials to the UAA server
  • The UAA validates the user and returns a valid JWT

Trusted Service to Service communication: Client Credentials grant type

The trusted service can request an access token using only its client-id and client-secret. When the client is requesting access to the protected resources under its control, it is very important that the client credentials grant type MUST only be used by confidential clients.

  • Zuul authenticates with his client-id and client-secret
  • The UAA validates the credentials and returns a valid JWT

OAuth2 Scopes

OAuth 2.0 scopes provide a way to limit the amount of access that is granted to an access token. If the scope is not defined, the client is not limited by scope.

An access token issued to a client can be granted READ or/and WRITE access to protected resources. If you enforce a WRITE scope to your API endpoint and it tries to call the endpoint with a token granted a READ scope, the call will fail

JSON Web Tokens (JWT)

JSON Web Tokens (JWT) is a compact URL-safe means of representing claims to be transferred between two parties. The claims in a JWT are encoded as a JavaScript Object Notation (JSON) object that is used as the payload of a JSON Web Signature (JWS) structure or as the plaintext of a JSON Web Encryption (JWE) structure, enabling the claims to be digitally signed or MACed and/or encrypted. The suggested pronunciation of JWT is the same as the English word “jot”. The payload consists of some standard attributes (called claims), such as issuer, subject (the user’s identity), and expiration time. The specification allows these claims to be customized, allowing additional information to be passed along. Be careful when passing additional information, if you like to go deeper on this topic with a real use case, you can read Using JWT for State Transfer Jwt.io provides a quick way to decode your JWT.

One of the challenges in a microservice-based architecture is identity propagation. After the authentication, the identity of the user needs to be propagated to the next microservice in a trusted way.
JWT is used here to carry along information of the user. Based on a token, your microservice needs to be able to create a principal object. This principal object needs to contain all the necessary info so the system can decide whether or not the request should be executed or not.


Dealing with time

When propagating the identity of the user, you don’t want it to last for a infinite amount of time.
That’s why JWTs have an expiration time. When expired, the JWT will be invalid and the client needs to request a new JWT with the refresh token. These refresh tokens carry the needed information to issue a new JWT. Refresh tokens can also expire but are rather long-lived. JWTs have three fields that relate to time and expiry, all of which are optional. In most cases, you should include these fields and validate that the token:

  • is not expired (exp)
  • was created before the current time (iat)
  • should not be used before the current time (nbf)

All of these times are expressed as UNIX epoch timestamps, and are best checked in the order as described above.

Signed JWTs

Signing a JWT helps establish trust between services, because it gives a recipient reason to believe that the message was created by a known sender and that the message was not altered in transit. JWTs are being signed by a public/private key pair. Almost all of the JWT libraries support signing. To check if yours supports it, visit JWT Libraries. For a deeper dive into signing JWT, check our tech post about Digitally signing your JSON Documents

  • The user requests a resource
  • The frontend assembles a request with an Authorization header and a Bearer token inside, fires off the request to Zuul
  • Zuul verifies the token in communication with the UAA server
  • If the token is valid, Zuul redirects the frontend to the correct resource on the proper microservice
  • The microservice checks for authorization to the resource, if access granted, the correct resource is returned


Since we are working with cloud-native applications, we can’t have any state within them. Because we have all the necessary information and create a new principal object for each request, the token eliminates the risk of having in-memory session state in the microservice.

Using a User Authentication & Authorization Server (UAA)

The UAA server is an identity provider. It adds authentication to applications and secures services with minimum fuss. It’s primary role is that of an identity provider, issuing tokens for client applications to use when they act on behalf of users. It can also authenticate users with their credentials, and can act as an SSO service using those credentials.

There are some options available as a UAA server:

Enabling Single Sign-On

Now that we have a way to achieve Authentication and Authorization by applying OAuth2 and JWT, we still have one problem. Having multiple frontends in our architecture, the user will have to log in to each of these applications. With Single Sign-On (SSO) we can eradicate this problem just by using the existing user session and requesting an access token.

Enable OAuth2 SSO flow on Zuul service

The @EnableOAuth2Sso and @EnableZuulProxy annotation on our Zuul service will forward OAuth2 tokens to the services it is proxying.

Sensitive Headers

Zuul secures your sensitive headers by blocking these headers downstream (microservice). Since the default settings for sensitive headers blocks the Authorization header, we have to open this setting and send these headers downstream. You can choose to set the sensitive header per route or globally.

How it works: Sensitive Headers

Zuul Filter

Based on Netflix’s Zuul, Spring’s implementation also brings a filter mechanism. Filters are capable of performing a range of actions during the routing of HTTP requests and responses. This can help you customize security on your incoming and outgoing traffic. Review the Zuul filter guide from Netflix about how filters work.

Securing your microservice

When enabling security in your service, the most common issues are developer-induced. Either there is a lack of built-in or easy security controls, or we make trade-offs for functionality over security. Still, we have to think about who can access this functionality and what they can do with it.

We got an access token, our gateway performed a coarse grained verification and proxied it to our microservice. We are in a ‘downstream service’, where data is being load-balanced from Zuul. The next questions are:

  • How do we decode this JWT?
  • How can we secure our code with the help of Spring Security?

Assembling the Principal

It is the responsibility of a microservice (Resource Server) to extract information about the user from the access token. Decoding the token allows the extraction of the user’s information. With this information Spring Security will assemble a Principal object containing eg. the username and the user’s roles, and puts it in the security context. Using the security context the AccessDecisionManager will be able to make a decision whether or not the request should be performed. To enable this, we need to add spring security to our class path and add the @EnableResourceServer annotation to our application.

Best practices with keys

The problem that might occur is that every microservice would need to connect with the UAA server for verification on every request.

Zuul verification

Obviously, we don’t want every microservice to depend on the UAA servers availability regardless of startup / testing / CI. The solution is to disable exposure of your microservices to the outer network and handle only incoming traffic via the gateway (eg. Zuul, HAProxy, nginx,…). Zuul will verify the token as a trustworthy client of the UAA server and will propagate the token to the downstream services. But what if a hacker gets inside of your platform?

JSON Web Keys

To solve this issue, we need an extra validity check on the microservice. When verifying a token’s validity, it comes down to verifying if the token was issued by the UAA server. This can be done by requesting the public key used for signing the JWT. This is called a JWK or JSON Web Key. Basically, you can restrict the dependency on the UAA server to one single REST call, where the JWK is fetched from a public URI. Once a microservice has a cached JWK, it can be used to verify any JWT completely by itself. This greatly reduces network calls to the UAA server and still secures all of your microservices. When you want to rotate your private/public key pair, you can use JWKS. We will go deeper into detail in one of our next posts.

Securing API endpoints

At last we’re going to secure our resources. Spring Security gives us a variety of tools to secure your application at class and method level. The one that’s used most often enables method security, which you enable by adding @EnableGlobalMethodSecurity(prePostEnabled = true) to your configuration.


For the authorization, Spring Security provides us with authorities, extracted from the access token. The authorities are placed inside a Principal, which will be used throughout the existing security context of your application. You can then reference them using Spring Expression Language (SpEL) to secure your methods. There are plenty of options you can use for method security, but we’ll highlight the most common ones.

You can find a complete list in the Spring documentation


Most commonly used, @PreAuthorize will decide whether a method can actually be invoked or not.

When a user logs in and you want the user to only access his detail information, or everyone’s data in case he’s an admin, you can use the @PreAuthorize annotation.

@PreAuthorize("(authentication.principal.uuid == #uuid.toString()) or hasRole('ADMIN')")
User findByUuid(@Param("uuid") UUID uuid);

Less commonly, you may wish to perform the access-control check after the method has been invoked. The returnObject is the returned value of that method.

A user can only view his own details and not those of someone else, but an administrator can. You validate this by checking if the user has the admin role or if the principal’s UUID is the same as the one of the returned user object.

@PostAuthorize("returnObject!=null or hasRole('ADMIN') or returnObject.uuid.toString() == authentication.principal.uuid")
User findOne(@Param("uuid") UUID uuid);

Next step

In the next post we will cover how to secure your data at rest.


Kevin Van Houtte is a Software Engineer at Ordina Belgium. Passionate in the Spring ecosystem and Angular framework Kevin is eager to discover new and efficient ways to solve problems. He enjoys a good challenge and is interested in cutting edge technologies. Currently he works on a project with Microservices, Spring and Angular/TypeScript.