I have a multi tenant Web app / API registered in Azure ad and that connected to an API App. The API App has Active Directory Authentication setup. At the moment only one other tenant needs access to api. I made sure only they can get access by putting https://sts.windows.net/<third party tenant>/ in the Issuer URL. My question is: How would I go about giving a second (or more) tenants access to the api? I can't add anymore tenant ids in the Issuer URL so I'm kinda at a loss
Thanks
The approach you are using currently will work only in a single tenant scenario, i.e. Automatic validation of tenant by setting IssuerURL works only in a single tenant scenario.
In case of multi-tenant applications, the application is responsible for storing and validating all possible issuers. This is by design and exact guidance on this topic from Microsoft is available here:
Work with claims-based identities in Azure AD: Issuer Validation
For a single-tenant application, you can just check that the issuer is
your own tenant. In fact, the OIDC middleware does this automatically
by default. In a multi-tenant app, you need to allow for multiple
issuers, corresponding to the different tenants. Here is a general
approach to use:
In the OIDC middleware options, set ValidateIssuer to false. This turns off the automatic check.
When a tenant signs up, store the tenant and the issuer in your user DB.
Whenever a user signs in, look up the issuer in the database.If the issuer isn't found, it means that tenant hasn't signed up. You
can redirect them to a sign up page.
You could also blacklist certain tenants; for example, for customers that didn't pay their subscription.
So, in case of a .NET based web application the code in your startup class would change to something like this.. notice the new TokenValidationParameters { ValidateIssuer = false }
Authenticate using Azure AD and OpenID Connect
app.UseOpenIdConnectAuthentication(new OpenIdConnectOptions {
ClientId = configOptions.AzureAd.ClientId,
ClientSecret = configOptions.AzureAd.ClientSecret, // for code flow
Authority = Constants.AuthEndpointPrefix,
ResponseType = OpenIdConnectResponseType.CodeIdToken,
PostLogoutRedirectUri = configOptions.AzureAd.PostLogoutRedirectUri,
SignInScheme = CookieAuthenticationDefaults.AuthenticationScheme,
TokenValidationParameters = new TokenValidationParameters { ValidateIssuer = false },
Events = new SurveyAuthenticationEvents(configOptions.AzureAd, loggerFactory),
});
Once you have disabled the Validate issuer, you will need to handle the validation yourself. Here is a sample with some guidance around how to do this validation yourself
Update your code to handle multiple issuer values
You will at least need to check the "tid" claim which captures the Azure AD Tenant Id against your own list of valid tenant IDs, before you let the call go through.
When a single tenant application validates a token, it checks the
signature of the token against the signing keys from the metadata
document, and makes sure the issuer value in the token matches the one
that was found in the metadata document.
Since the /common endpoint doesn’t correspond to a tenant and isn’t an
issuer, when you examine the issuer value in the metadata for /common
it has a templated URL instead of an actual value:
https://sts.windows.net/{tenantid}/
Therefore, a multi-tenant application can’t validate tokens just by
matching the issuer value in the metadata with the issuer value in the
token. A multi-tenant application needs logic to decide which issuer
values are valid and which are not, based on the tenant ID portion of
the issuer value.
For example, if a multi-tenant application only allows sign in from
specific tenants who have signed up for their service, then it must
check either the issuer value or the tid claim value in the token to
make sure that tenant is in their list of subscribers. If a
multi-tenant application only deals with individuals and doesn’t make
any access decisions based on tenants, then it can ignore the issuer
value altogether.
(EDIT) More information on Validating Tokens
I'm trying to answer your questions from comments here.
Here is sample code which does the task of manually validating JWT tokens.
Manually validating a JWT access token in a web API
A useful excerpt..
Validating the claims When an application receives an access token
upon user sign-in, it should also perform a few checks against the
claims in the access token. These verifications include but are not
limited to:
audience claim, to verify that the ID token was intended to be given
to your application not before and "expiration time" claims, to verify
that the ID token has not expired issuer claim, to verify that the
token was issued to your app by the v2.0 endpoint nonce, as a token
replay attack mitigation You are advised to use standard library
methods like JwtSecurityTokenHandler.ValidateToken Method
(JwtSecurityToken) to do most of the aforementioned heavy lifting. You
can further extend the validation process by making decisions based on
claims received in the token. For example, multi-tenant applications
can extend the standard validation by inspecting value of the tid
claim (Tenant ID) against a set of pre-selected tenants to ensure they
only honor token from tenants of their choice.
Sample Access Token, just for understanding: Access Token and Id_token are both simple base64 encoded JSON Web Tokens (JWT). You can decode these to find the claims and then validate them. I'm sharing a sample which has code to do just that. Before that here is a sample access token from one of Microsoft Docs. I just took one for example from here
Actual Value: eyJ0eXAiOiJKV1QiLCJhbGciOiJSUzI1N... (a long encoded string continues)
Decoded Value (you can check this easily using a website like https://jwt.io):
{
"aud": "https://service.contoso.com/",
"iss": "https://sts.windows.net/7fe81447-da57-4385-becb-6de57f21477e/",
"iat": 1388440863,
"nbf": 1388440863,
"exp": 1388444763,
"ver": "1.0",
"tid": "7fe81447-da57-4385-becb-6de57f21477e",
"oid": "68389ae2-62fa-4b18-91fe-53dd109d74f5",
"upn": "frankm#contoso.com",
"unique_name": "frankm#contoso.com",
"sub": "deNqIj9IOE9PWJWbHsftXt2EabPVl0Cj8QAmefRLV98",
"family_name": "Miller",
"given_name": "Frank",
"appid": "2d4d11a2-f814-46a7-890a-274a72a7309e",
"appidacr": "0",
"scp": "user_impersonation",
"acr": "1"
}
As you can see the decoded value has many claims including "tid" which you're about to validate.
Hope this helps!
Related
We're implementing a custom identity provider for Azure AD B2C, using OpenID protocol option, as a generic OpenID Connect.
Everything works as expected until it's time to post the response back to Azure AD B2C using the redirect URI provided. I've found documentation regarding expected structure of this response URL, and what we see in the documentation is identical to what Azure AD B2C specifies when it issues the authentication sequence.
Configured values:
Response type: code
Response mode: form_post
User ID claim: sub
Display name claim: name
When the custom identity provider GETs or POSTs authentication response (code) back to https://REDACTED.b2clogin.com/REDACTED.onmicrosoft.com/oauth2/authresp, the Azure B2C returns 404.
Note that this is not 400, not 401, not 403, not 5xx. It is precisely 404 (not found), with a basic text (non-html) content saying resource not found. This response looks to me very much like a misconfigured API management layer on Azure side, hitting a wrong internal URL.
We're expecting that the URL https://REDACTED.b2clogin.com/REDACTED.onmicrosoft.com/oauth2/authresp actually works. It looks like what the expected Azure AD B2C response endpoint is from documentation, and it is also exactly what Azure AD B2C itself specifies when initiating the OpenID sequence with our custom identity provider web application.
So far we were unable to find the root cause, nor even any useful input beyond raw network request logs (case with Microsoft support was open since 2023-01-23). The last resort could be re-creating the B2C tenant, since this feature seems to work for other people, but that would require migration and significant down time on our end.
SOLUTION: The response to AD B2C authresp endpoint was missing 'nonce' claim (in the id_token payload), and 'state' parameter in the HTTP request. Both values are supplied by AD B2C when initiating authorization. As soon as custom identity provider started properly adding those two values, error 404 went away.
Response should include supplied nonce as a claim inside the id_token payload, and supplied state as HTTP request parameter or query string
https://openid.net/specs/openid-connect-basic-1_0.html
I had the same issue (a 404 error as a result of the /authresp POST from my custom OIDC IdP back to Azure AD B2C using the redirect URI Azure AD B2C had just provided as a query parameter on the /authorize request to my IdP: redirect_uri=https://mytenant.b2login.com/mytenant.onmicrosoft.com/oauth2/authresp
In my case (using an implicit flow), it was about properly handling the "nonce" query parameter on the inbound /authorize request (from Azure AD B2C to my IdP) by ensuring the generated id_token it returned included the nonce as a claim.
In your case (using an authorization code flow...and assuming you also return an id_token based on the "sub" and "name" claims you're returning), your /token endpoint needs to include the nonce inside the id_token...so propagate the nonce (and state) as query parameters along to your /token endpoint via the /authorize to /token redirect method you use.
If a federated IdP doesn't include the nonce as a claim inside the id_token payload that it returns, Azure AD B2C will return a 404 error from the /authresp request.
I don't know why Microsoft chose to return a 404 instead of a more informative "nonce invalid" error message, or at least, a 400 error...perhaps it's for the same security reason a login form doesn't precisely tell you when your password is invalid.
In the OpenID Connect specification, the nonce description (under IDToken) states (bolding is my doing):
String value used to associate a Client session with an ID Token, and to mitigate replay attacks. The value is passed through unmodified from the Authentication Request to the ID Token. If present in the ID Token, Clients MUST verify that the nonce Claim Value is equal to the value of the nonce parameter sent in the Authentication Request. If present in the Authentication Request, Authorization Servers MUST include a nonce Claim in the ID Token with the Claim Value being the nonce value sent in the Authentication Request. Authorization Servers SHOULD perform no other processing on nonce values used. The nonce value is a case sensitive string.
Although the spec indicates a nonce is optional, Microsoft is following best practices by supplying one...and since Azure AD B2C (as the Authorization Server) gets the id_token from the IdP, it requires a federated OIDC IdP to play by the same rule.
In case this helps others, my custom IdP's /.well-known/openid-configuration endpoint returns:
{
"authorization_endpoint": "https://myidp.azurewebsites.net/oauth2/authorize",
"authorization_response_iss_parameter_supported": true,
"claims_parameter_supported": false,
"claims_supported": [
"aud",
"idp",
"iss",
"iat",
"exp",
"nonce",
"s-hash",
"sid",
"sub",
"auth_time",
"email",
"family_name",
"given_name",
"locale",
"name",
"updated_at",
"user_id"
],
"claim_types_supported": ["normal"],
"grant_types_supported": ["implicit"],
"id_token_signing_alg_values_supported": ["RS256"],
"issuer": "https://myidp.azurewebsites.net",
"jwks_uri": "https://myidp.azurewebsites.net/oauth2/jwks",
"response_modes_supported": ["form_post"],
"response_types_supported": ["id_token"],
"scopes_supported": ["openid"]
}
(Yes, my IdP runs on an Azure App server...but, "myidp" isn't my real tenant name.)
p.s. Currently, my IdP is used exclusively in a federation with AzureAD B2C (which acts as the Authorization Server for my client application via the MSAL library), so my IdP simply supports just an implicit flow and three endpoints (/.well-known/openid-configuration, /jwks and /authorize). If it were a general purpose IdP, or allowed direct client requests, it would support other flows (e.g. an authorization code flow), additional scopes (beyond "openid"...e.g. "profile") and additional endpoints (e.g. /token and /userinfo). However, regardless of flow, as long as an id_token is returned, it needs to include the nonce in its payload.
To troubleshoot the issue, I would recommend the following steps:
Verify that the redirect URI you are using is correct and matches
the one specified by Azure AD B2C.
Check that the response type and response mode specified in your
custom identity provider match the values expected by Azure AD B2C.
Verify that the claims you are sending in the response (e.g. "sub"
and "name") match the expected format and values for Azure AD B2C.
Check the network request logs for any additional information that
might help identify the issue.
If possible, try to isolate the issue by testing the authentication
flow with a minimal configuration to determine if the problem is
with your custom identity provider or with Azure AD B2C.
If the issue persists after trying these steps, you may want to consider reaching out to Microsoft support for further assistance.
I tried to reproduce the scenario in my environment:
Make sure the endpoint to which I requested the authorization url
It includes policy and with
redirect URI= https://kavyasarabojub2c.b2clogin.com/kavyasarabojub2c.onmicrosoft.com/oauth2/authresp
User Flow is of SignupSignin and not just Signin
Make sure to include all the required api permissions , importantly make sure to include openid , profile
I Configure idp such that , userId is mapped to oid.
The authorization url must have the policy included .
Here I have B2C_1_SignupSignin policy set for the User flow.
redirect URI= https://kavyasarabojub2c.b2clogin.com/kavyasarabojub2c.onmicrosoft.com/oauth2/authresp
Auth url:
https://kavyasarabojub2c.b2clogin.com/kavyasarabojub2c.onmicrosoft.com/oauth2/v2.0/authorize?p=B2C_1_newSignupSignin&client_id=xxx&nonce=defaultNonce&redirect_uri=https%3A%2F%2Fxxxb2c.b2clogin.com%2Fxxxb2c.onmicrosoft.com%2Foauth2%2Fauthresp&scope=openid&response_type=id_token&prompt=login
When profile scope is not given I got bad request
But when openid and profile along with Directory.Read.All api permissions are included, the request run successfully.
Note: metadata url must be : https://login.microsoftonline.com/<tenantId>/v2.0/.well-known/openid-configuration
Successfully logged in and got the token containing idp_access_token
Identity provider access token , decoded and got the user claims:
I am configuring an OIDC-based SSO flow in Azure AD B2C using custom policy to allow users to login to downstream applications with their federated identity provider's (IdP) credentials. Custom policy is used to allow some complex business logic to be run prior to providing the token to the downstream applications.
The flow is correctly redirecting users to the external IdP for login and ultimately back to my downstream applications with associated claims. However, there is a custom claim that is only available in the access token received by B2C from the external IdP (not the ID token), and I can't figure out how to retrieve this claim from the access token to be used in the B2C user journey and ultimately provided with all the other claims to the downstream applications.
I can see that B2C does receive both the ID and access tokens by reviewing Application Insights logs (sample output):
"TESTtechnicalprofile": {
"ContentType": "Jwt",
"Created": "2022-10-15T07:37:45.8678974Z",
"Key": "TESTtechnicalprofile",
"Persistent": true,
"Value": "eyJhb..."
},
"TESTtechnicalprofileaccess_token": {
"ContentType": "Unspecified",
"Created": "2022-10-15T07:37:45.8678974Z",
"Key": "TESTtechnicalprofileaccess_token",
"Persistent": false,
"Value": "eyJhb..."
},
And general format of the payloads of the tokens is as follows:
ID Token
{
"<name of custom claim I can retrieve>": "custom claim value",
"iss": ...,
...
}
Access Token
{
"<name of custom claim I cannot retrieve>": "custom claim value",
"iss": ...,
...
}
I can successfully retrieve claims from the ID token by mapping from the partner claim type:
<OutputClaim ClaimTypeReferenceId="email" PartnerClaimType="<name of claim containing email in ID token>" />
However this same method doesn't work for claims in the access token. If I reference the name of the custom claim in the access token in PartnerClaimType then B2C omits the claim (presumably because it fails to retrieve it).
I've tried retrieving the access token itself as a claim using the method described here and that works (token in claim matches token seen in Application Insights logs), however I'm not sure if it's possible to decode this token in B2C policy and subsequently pull claims from it (or even if one would want to do that).
While I could let the downstream applications retrieve what they need from the access token, I have business logic in my user journey that needs this claim prior to providing the final token to the applications.
Following up here for anyone else trying to do the same, according to Microsoft Support it isn't possible to extract a claim from an access token in B2C policy. I ended up crafting a workaround involving calling an external REST API from B2C to retrieve the needed info for the user journey.
In JWT from AAD there is a key 'aud'. https://jwt.io/, says it is 'Audience. (Who or what the the token is intended for)'. My question is, Are aud values website specific - can I check the aud and expect it to be same to check if the token is intended for my specific site?
In Azure AD, the audience value always indicates the resource the token is targeted on.
You can acquire an access token by using either the API's client id or Application ID URI.
What you use will be the audience in the token.
So if you make an API, you should check the audience is either the API's client id or Application ID URI.
You can know for sure it will always be one of those if the token is meant for your API.
EDIT: The below information is not correct.
If I know your API's identifier + your tenant id,
I can acquire an access token for your API using client credentials!
The token will not contain scopes or roles, it cannot.
So it is critical that you check for the presence of valid delegated permissions (aka scopes) or valid app permissions (in roles claim).
THIS IS WRONG: If I tried to acquire an access token using your API's identifier from my AAD tenant, it would not give me a token.
Any app that passes an access token with the correct audience had rights to call your API when it acquired the token.
You already got a good explanation of the audience value from juunas.
I'm adding here a specific code example from Azure-Samples on Github which shows how to validate the JWT Token manually and checks among other things audience value. (It's pretty important to validate issuer as well)
Look at this particular code and especially near the comment "We accept both the App Id URI and the AppId of this service application"
JwtSecurityTokenHandler tokenHandler = new JwtSecurityTokenHandler();
TokenValidationParameters validationParameters = new TokenValidationParameters
{
// We accept both the App Id URI and the AppId of this service application
ValidAudiences = new[] { audience, clientId },
// Supports both the Azure AD V1 and V2 endpoint
ValidIssuers = new[] { issuer, $"{issuer}/v2.0" },
IssuerSigningKeys = signingKeys
};
Code Sample:
Specific file with code excerpt shown above
Azure-Samples: Manually validating a JWT access token in a web API
I have create a webapi secured with azure active directory. I need to test this now and trying to use fiddler with an authorization header. I am trying to generate the token with below code.
Target obj = (Target)cmbTarget.SelectedItem;
AuthenticationResult authenticationResult;
string aadInstance = obj.AADInstance; // "https://login.windows.net/{0}";
string tenant = obj.Tenant; //"rudderless.onmicrosoft.com";
string apiResourceId = obj.ApiResourceId; //"15b4ac7f-23a8-4958-96a5-64159254690d";
string clientId = obj.ClientId; // "47cdc6c3-226a-4c38-b08e-055be8409056";
Uri redirectUri = new Uri(obj.RedirectUri); //new Uri("http://nativeclient");
string authority = string.Format(aadInstance, tenant);
authContext = new AuthenticationContext(authority);
authenticationResult = this.authContext.AcquireToken(apiResourceId,
clientId, redirectUri, PromptBehavior.Always);
txtToken.Text = authenticationResult.AccessToken;
Clipboard.SetText($"Bearer {txtToken.Text}");
I get the token generated successfully and when I am using the token to call the webapi it throwing 401 with message
WWW-Authenticate: Bearer error="invalid_token", error_description="The
audience is invalid"
I think it is important to revisit the different steps of authentication, and hopefully through the discussion you will be able to solve the issue you are having.
When a client is trying to get an access token to a resource, it needs to specify to AAD which resource it wants to get a token for. A client may be configured to call multiple resources, all with different configurations, so it is an expectation that the resource is always specified in an Access Token Request.
The resource can either be an App ID GUID for the Resource, or a valid App ID URI which is registered on the Resource. AAD should be able to uniquely identify which resource you are trying to reach based on the value you provide. However, note that if you use an App ID GUID, you will get a token from AAD where the Audience claim is the App ID GUID. Alternatively, if you use an App ID URI, you will see that URI as the audience claim in the token.
In both situations, you will get a token for the 'same' resource, but the claim in the token will appear differently. Additionally, it may be possible that a single application resource may have multiple App ID URIs registered on their app. Depending on which one you use in the authentication request, you will get a different audience claim in the token which matches the resource parameter you passed in.
Finally, once you get the token, you send it over to the Resource API who will validate the token for a number of things, such as: the Client ID Claim, the Scopes/Roles Claims, the authentication method ('acr' claim), and definitely that the audience claim matches what they expect!
This means that the Resource API ultimately needs to say "I accept < App ID GUID > as a valid Audience Claim"... or "I accept < App ID URI > as a valid Audience Claim". This kind of logic may be built into the library you are using (like OWIN), but you need to make sure that on your API side, you have it configured correctly for the Audiences you expect. You could, if you wanted, make it so that your API does not check the Audience claim at all! All the claims in the token are plaintext, and thus you could really do whatever you want, but you would not have a very secure API in that situation :]
End of the day, my hunch is that this error is coming from your own API, and it is happening because you have not configured your app to accept an Audience claim which matches your Resource's App ID GUID (which it looks like what you are passing when you are getting a token based on your code sample).
I hope this solves your issue!
Problem
After implementing the instructions found in this Protected web API: Code configuration article, I received an error message similar to the OP's:
WWW-Authenticate: Bearer error="invalid_token", error_description="The
audience is invalid"
The problem turned out to be my AzureAd > ClientId setting in my appsettings.json file.
Solution
I updated the appsettings.json file of my ASP.NET Core Web API app so that the ClientId setting used the "Application ID URI" found in portal.Azure.com under my App Registriation > "Expose An API" section.
The section in appsettings.json looks similar to this:
"AzureAd": {
"Instance": "https://login.microsoftonline.com/",
"TenantId": "XXXXXXXX-XXXXX-XXXXX-XXXXX-XXXXXXXXXX",
// ClientId = Portal.Azure.com > App Registration > Expose an API > "Application ID URI"
"ClientId": "api://XXXXX-XXXXXX-XXXXX-XXXX-XXXXXXXXX"
}
Important note
"aud" value that is being generated for JWT token by azure is also controlled by "accessTokenAcceptedVersion" property in AD application manifest.
This property defines a version of the access token that will be generated (MS docs about accessTokenAcceptedVersion).
Possible results for its values:
null or 1 - "api://" prepended to GUID
2 - "api://" is not added, so there should be GUID only
I had the same issue. Thought of sharing it. I have change the Web Api Audience to the ClientId of the Web App. After this it works.
The Microsoft references show the following example:
{
"AzureAd": {
"Instance": "https://login.microsoftonline.com/",
"ClientId": "[Client_id-of-web-api-eg-2ec40e65-ba09-4853-bcde-bcb60029e596]",
"TenantId": "common",
"Audience": "custom App ID URI for your web API"
},
// more lines
}
Can also be that your app/lib is using a newer version of the api.
If accessTokenAcceptedVersion is null in the manifest of your app ms defaults to v1.
Check your jwt token in http://jwt.io
If you get this - check your JWT Token. If ISS isn't like this
"iss": "https://login.microsoftonline.com/[yadyada]/v2.0",
then most likely you're using another version (like version 1 which is default). Check the manifest of your azure ad app:
Below value is probably null or one, should be two:
"accessTokenAcceptedVersion": 2,
I had the same issue. I was using the client's Resource ID as the parameter for AcquireToken when I should have used the server's Resource ID.
It works when I use the correct Resource ID.
I got the same error. It was because I was using a custom domain, so my API ID URL wasn't api://{client-id}.
The solution is to set the Audience setting on your appsettings.json, just like mentioned in the Microsoft Wiki:
"AzureAd": {
"Instance": "https://login.microsoftonline.com/",
"TenantId": "XXXXXXXX-XXXXX-XXXXX-XXXXX-XXXXXXXXXX",
"ClientId" : "XXXXXXXX-XXXXX-XXXXX-XXXXX-XXXXXXXXXX",
// Audience = Portal.Azure.com > App Registration > Expose an API > "Application ID URI"
"Audience": "Application ID URI"
}
While calling api for implementing service principle through App registration in active directory.
I got this error while calling api-GET {vaultBaseUrl}/secrets/{secret-name}/{secret-version}?api-version=7.0 with bearer key to get key vault secret value.
As part of fix, to get bearer value, Apart from passing clientid, client secret, grant_type,I added resource key with value https://vault.azure.net as part of request body of api call for https://login.microsoftonline.com/{ActiveDirectoryId}/oauth2/token.
This might help someone: I've encountered this error because the MS Graph User.Read permission was missing on the SharePoint Online Client Extensibility Web Application Principal. Out of the box, this app reg already has the User.Read permission, but I had removed that one because (for an earlier project) I already used User.Read.All, thinking that it included User.Read. However, User.Read is used for sign-in purposes while User.Read.All is not. When I restored User.Read, my problem was solved.
Quite the unintuitive solution.
I'm considering using the JWT audience field to implement role-based authorization in my app.
So I'd have ServiceA which requires 'RoleA' audience to be present, ServiceB requires 'RoleB' etc. Then when I issue the JWT, I include the appropriate audience(s).
Relevant section from the JWT draft spec:
The aud (audience) claim identifies the recipients that the JWT is intended for. Each principal intended to process the JWT MUST identify itself with a value in the audience claim. If the principal processing the claim does not identify itself with a value in the aud claim when this claim is present, then the JWT MUST be rejected... The interpretation of audience values is generally application specific.
So it appears that would work but since I'm new to JWT I'm wondering: is role-based authorization an appropriate use case for the audience field? Or should I roll my own logic using a payload with custom roles array etc?
Thanks
I understand audience rather then list of consumers/applications who can authorize the user.
In my application I put roles into own array in the payload. For example like that.
{
"sub": 1234567890,
"exp": 9876543210,
"name": "John Doe",
"roles": ["USER", "EDITOR"]
}
On the server I am authorized using spring security and user loaded from "sub".
And on the client I can use these roles to show proper buttons and fields.