We are current building a collection of back-end ASP.NET Core microservices. These services will not be accessed directly from the front-end application, but rather accessed through an ASP.NET Core API gateway. We are using IdentityServer4 for the OpenID Connect server. I have been able to setup the UseJwtBearerAuthentication middleware to have API gateway validate the JWT bearer token (access_token) against IdentityServer4. I would like to be able to have the API gateway inject the id_token, based on the access_token, into the requests made to the back-end services that may need to know the end-user.
Is there a way to configure the JWT middleware to retrieve the id_token when validation the access_token or do I need to manually call the OpenID Connect server in the API gateway?
You don't use id_tokens at APIs - they are for clients.
If you want to have access to certain identity claims, either include them in the access token (by configuring the ScopeClaims on the resource scope), or use the access token to contact the userinfo endoint which in turn will return the identity claims.
The JWT middleware performs standalone verification, it does not contact the identity server to verify or retrieve anything. You'll have to make an additional call.
Related
I'm learning identityserver4 and can't understand some stuff.
Shortly: I want to authorize end-users with email and password on ASP.NET MVC client side (it will send user credentials to the token server to get tokens), and I don't want third-party clients to retrieve data from my API resources.
As I understood from the documentation:
implicit flow is used for SPA (js clients) and uses id_token to authorize users. I can store id_token in my browser.
client credentials flow is used for trusted apps (like ASP.NET MVC client) to authorize clients and uses access_code. I can store access_code inside my app.
So looks like I need a hybrid flow.
In the documentation I read that I need to use AddOpenIdConnect() method and
Technically the tokens are stored inside the properties section of the cookie.
So my questions:
If the id_token can be stored in the browser, why is it not safe to store the access_token there too ?
As the docs state, the tokens are stored inside the properties section of the cookie. It's confusing, because some guides says, that it's not safe to store access_token there. So where I should store access token in my ASP.NET MVC client?
Am I right, that AddOpenIdConnect() configures my ASP.NET MVC app to retrieve access_token automatically from the token server? And if yes - in which moment should I authorize users with email/password and how to combine all tokens inside one JWT on my ASP.NET MVC client's backend when I will send requests to api resources ?
Today you should not use the Implicit Flow and it has been deprecated as of Oauth 2.1. What you should use is the authorization code flow with PKCE. PKCE is a security enhancement for the authorization code flow.
So as of OAuth 2.1 you only have two main flows:
Authorization code flow, for a MVC client to login a user to the client
Client credentials flow, for API->API communication where no human is involved.
To answer your questions:
If the id_token can be stored in browser, why it's not safe to store access_token there too ?*
The ID-Token is only used to create the initial user session and after that you can throw it away. It also only have a short life time of 5 minutes by default in IdentityServer.
As the docs state, the tokens are stored inside the properties section of the cookie. It's confusing, because some guides says, that it's not safe to store access_token there. So where I should store access token in my ASP.NET MVC client?
The tokens can be stored in your session cookie in ASP.NET Core and that's secure. It's protected / encrypted using the Data Protection API. However, the cookies can grow in size quite a lot if you do that.
Am I right, that AddOpenIdConnect() configures my mvc app to retrieve access_token automatically from the token server? And if yes - In which moment I should authorize users with email/password and how to combine all tokens inside one JWT on my ASP.NET MVC client's backend when I will send requests to api resources?
AddOpenIdConnect only handles the initial login and retrieving the first ID and access token. It does not handle refreshing of the access token using refresh tokens. For that you can add the IdentityModel library.
Today when you use the auth code flow, you redirect the user to IdentityServer and you let the user login there instead of passing the username/pwd from you browser to identityserver.
I'm trying to configure authentication and authorization middleware in an asp.net core 3.1 API project to be able to authorize users from:
on-premise IdentityServer4 (IDS) and
AD managed in Azure (AAD).
I'm planning to use JWT bearer tokens in both cases to call the API endpoints.
The IDS users are authenticated in a mobile app against the IDS.
The AAD users are authenticated in an admin SPA against the AAD.
The web api has separate public endpoints for IDS users and management endpoints for AAD users.
There are lots of working samples about how to configure web APIs for IDS and AAD separately, but not together.
Does anyone have an example with how this should be done?
Is it even possible to use different auth schemes (e.g JwtBearerDefaults.AuthenticationScheme, OpenIdConnectDefaults.AuthenticationScheme) for an incoming JWT bearer token?
Any directions will be appreciated.
You can have multiple AddOpenIdConnect schemes, one for AzureAD and one for IdentityServer and then let the user challenge one of them, like:
await HttpContext.ChallengeAsync("identityserver");
or
await HttpContext.ChallengeAsync("azuread");
JwtBearerDefaults only deals with processing incoming tokens and creating a user out of it.
But each JwtBearer API's likes to have one "authority"-server (not multiple). So you might need to to add multiple AddJwtBearer(one for each server).
Or you use IdentityServer as your main authorization provider and then login to Azure through IdentityServer. So your internal servers only needs to know about IdentityServer and its tokens.
I have an external endpoint which is going to hit the Azure API gateway and that would route it to the backend API which is protected by IdentityServer4 authorization.
I am getting the access token if I hit it through the Postman client with the interactive UI from IdentityServer.
Is there a way I can get the access token required from the Azure API Management to validate against the IdentityServer4 and append it to the header in the request to the backend API?
Yes it is possible to achieve it through custom policy. You can ask your external API-Client/Consumer to paas in credentials in heaser, and then you write a policy inside inbound to can read those user credentials and do a API request (similar to your postman) and get the access token. You can then append the same token and let your request gets forwarded to backend API.
As per your problem statement, this should work. In case not, you might have to explain your scenario with more description/steps.
Here are some of the reference materials for you, I hope it helps.
https://learn.microsoft.com/en-us/azure/api-management/api-management-advanced-policies#SendRequest
https://learn.microsoft.com/en-us/azure/api-management/api-management-sample-send-request
Postman has a luxury of a human user seeing the UI and authorizing API access and IdentityServer4 to issue a token for Postman. There is no such luxury when call is being processed by APIM server, as you could send request for token to IdentityServer4, but who would be presented UI to authorize the action?
The only way is to provision some sort of secret to APIM (header, query, certificate) that would be recognized by IdentityServer4 to allow it issuing tokens for APIM. If such secred is available you could use send-request policy to make a call to IdentityServer4 and obtain required token.
Or make sure that every request to APIM has a token already.
we plan to introduce an API management solution and we're currently setting up a proof of concept with WSO2 AM. We want to use the WSO2 API gateway to check whether a certain consumer application is allowed to use an API and to throttle the request rate.
I work on the identity workflow and I wonder how a consuming application can pass a JWT token to the backend service with WSO2-AM in between.
First, this is our current scenario:
Without API gateway
The consuming application gets a JWT token for its carbon user from an identity provider. The JWT contains some claims about the user, e.g. the roles he/she belongs to.
The app calls the service an passes the JWT token in the Authorization HTTP header like: Authorization: Bearer
The service validates the issuer and signature of the JWT and retrieves the claims from it.
So, this is pretty straight forward. Now we put an API gateway in between the application and the service:
With API gateway
The consuming application gets a JWT token for its carbon user from an identity provider.
The consuming application uses OAuth2 to get an access token for the following API calls. We can use the client_credentials grant type and simply pass the the client id and client secret. I haven't yet tried it, but we could possibly use the JWT grant type (see https://docs.wso2.com/display/ISCONNECTORS/Configuring+JWT+Grant+Type) and use the JWT for passing user information to the API gateway.
The API gateway validates the JWT against the public key of the identity provider when using the JWT grant type.
An access token is returned to the app.
The app sends an API request to the gateway and passes the access token in the Authorization HTTP header.
The gateway validates the access token.
The gateway forwards the API request to the service.
And there is my problem: How can the JWT from 1/2. be passed to the service?
There is a documentation for "Passing Enduser Attributes to the Backend Using JWT" (see https://docs.wso2.com/display/AM210/Passing+Enduser+Attributes+to+the+Backend+Using+JWT), but this would introduce a new JWT, issued and signed by WSO2-AM, and I'm not sure, whether this JWT contains all information from the JWT used to create the access token (or even the original JWT).
Another way I could think of is using a custom HTTP header for passing the JWT through the gateway to the service. I cannot use the Authorization header (as we do without the API gateway), because WSO2-AM expects the access token in that header.
Since I'm not happy with either solutions, I want to ask the experts: How would you solve this?
Thanks,
Torsten
The only possibility I can think of is to send the JWT token in a custom Header for the backend service.
I've some questions regarding authentication in a microservices architecture. I've right now a monolithic application and my goal is to split the application in small microservices.
My bigest problem is for authentication (for now). After reading a LOT a documentation, It seems that the best solution is to use OpenID Connect to authenticate an user to retrieve a JWT that can by passed with the request to the microservices.
Also, to avoid having multiple endpoints, you can deploy and API Gateway to have only one endpoint for the end user. Ok, so now I've two questions with this architecture.
The standard flow for authentication will be :
An user contact my identity server in OpenID Connect with the implicit flow and get the id_token (JWT) and also the access_token. The user can now contact my API with this access_token. The API Gateway will valide the access_token with the identity server and also retrieve the JWT to add it to the sub request to the microservice API.
1/ How the API Gateway can get the JWT from the access_token? From what I red from the documentation (http://openid.net/specs/openid-connect-core-1_0.html), It can contact the "/userinfo" endpoint but It will get just the JSON format not the JWT...
2/ I want to allow authenticated calls between my microservices. So each microservice needs to be able to generate a JWT to contact other microservices directly. My first thought was to contact the identity server. But with the OAuth2 Client Credentials flow, I don't retrieve a id_token or a JWT. Just a classic OAuth2 access token without JWT. My second thought was that the microservice can directly sign its own JWT with a certificate issued by the same PKI as the one used by the identity server. That mean that a JWT can be sign by several certificats but from the same private PKI. When a microservice receives a JWT, It needs to be able to identify witch certificat was used to sign the JWT. I don't find anything on the RFC regarding this problem. I can add my own private claim in the token to have the certificate but after several days of browsing the web without seeing this kind of solution, I'm wondering if I'm not on the wrong path... To sum up, how can i perfom "User to service" authentication AND alors "service to service" authentication in JWT?
Thank you very much!
I am implementing a similar solution. Not sure if it will address to your question completely, but, I hope it helps:
You can implement a new authentication micro-service to convert your oAuth2 access token to JWT token. This microservice will also sign this JWT token.
Your API gateway will route all client requests to authentication service, which will validate this token from IDM and will convert it to a signed JWT token.
API gateway will pass this JWT token to other microservices which will validate the signature from Authentication Service's public key. If the signature validates, roles can be extracted out of it for authorization.
Each microservice can have its own IDM credentials configured and when it wants to call any other microservice, it can generate an access token and call Authentication Service to get JWT which can be passed in call to other microservices.