Just for a change, I was asked to help out on a non-Identity project.
There was a legacy host that only understood SOAP and a modern back-end that only supported REST web API.
So we need a bridge between them and I had to remember everything I had ever forgotten about WCF.
I needed my WCF method to call:
string response = await my_api.CallREST(parameter);
The problem is that the compiler expects the method to be decorated with async.
Aync. WCF?
Turns out you can.
My method looks like:
public async Task<validateresponse> ValidateParameter (parameter)
... and similarly for the interface.
And it works!
The word "async" doesn't appear anywhere is the WSDL. It appears to be completely ignored.
Enjoy!
Ideas and thoughts about Microsoft Identity, C# development, cabbages and kings and random flotsam on the incoming tide
Thursday, March 30, 2017
Friday, March 24, 2017
ADFS : Copy claims rules over
Don't know how many times I've done this.
Deploy to Dev., get everything working, deploy to QAS, support QAS acceptance testing, deploy to Prod., smoke test.
When you have a lot of claims rules to copy over, I found a neat way to do it.
e.g. for a CP.
(Get-AdfsClaimsProviderTrust -Name "My CP").AcceptanceTransformRules | Out-File “C:\path\CPClaimsRules.txt”
And then import the rules to the new CP.
Set-AdfsClaimsProviderTrust -TargetName "My CP" -AcceptanceTransformRulesFile “C:\path\CPClaimsRules.txt” .
I've found that this also avoids the issue where you keep a copy of the rules in e.g. Word and then when you try and paste them into the ADFS wizard, you get all kinds of format errors.
Update: But beware copying groups over.
Enjoy!
Deploy to Dev., get everything working, deploy to QAS, support QAS acceptance testing, deploy to Prod., smoke test.
When you have a lot of claims rules to copy over, I found a neat way to do it.
e.g. for a CP.
(Get-AdfsClaimsProviderTrust -Name "My CP").AcceptanceTransformRules | Out-File “C:\path\CPClaimsRules.txt”
And then import the rules to the new CP.
Set-AdfsClaimsProviderTrust -TargetName "My CP" -AcceptanceTransformRulesFile “C:\path\CPClaimsRules.txt” .
I've found that this also avoids the issue where you keep a copy of the rules in e.g. Word and then when you try and paste them into the ADFS wizard, you get all kinds of format errors.
Update: But beware copying groups over.
Enjoy!
Wednesday, March 22, 2017
ADFS : Creating a custom attribute store
This is for ADFS 4.0 on Server 2016.
This is a good write-up.
Unfortunately, all the code is in a screen shot which sucks. Somewhat difficult to copy / paste :-).
Luckily, similar code can be found here.
Just standardise the names; one is "ToUpper"; the other is "toUpper".
However, my requirement was for getting claims from a back-end (details unimportant for the purposes of this post) where a user could have many claims of that type returned. Think of a property ID where one person owns a house but an investor owns several.
All the examples were for returning one attribute.
Essentially, you are returning a C# jagged array e.g. string[][] resultData.
My query string was :
c:[Type == "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/emailaddress"]
=> issue(store = "CAS", types = ("HouseID"), query = "House", param = c.Value);
So I assumed that the multiple claims would be in the same row i.e. one row; many columns.
The search to the back-end returned 3 houses.
I then got:
Microsoft.IdentityServer.ClaimsPolicy.Language.PolicyEvaluationException: POLICY0019: Query 'House' to attribute store 'CAS' returned an unexpected number of fields: expected '1', got '3'.
If you look at the query, you'll see there is only one query parameter which will only return one result. A query which returns 3 results would be like:
query = "House", sn, mail
So what I need is 3 rows; one column.
I found some guidance around this here.
One of the things that stumped me for a while was the fact that the array had to be dynamic because there could be any number of houses. That's why you add to a list and then cast to an array.
Another gotcha was the fact that while you can have a static rule like:
=> issue(type = "HouseID", value = "123456");
you cannot have that in a query string for the attribute store. You get:
System.ArgumentException: ID4216: The ClaimType 'HouseID' must be of format 'namespace'/'name'.
So it needs to be:
c:[Type == "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/emailaddress"]
=> issue(store = "CAS", types = ("http://claim/HouseID"), query = "House", param = c.Value);
i.e. HouseID becomes http://claim/HouseID.
Once you have compiled a new .dll file, you have to copy it over in the ADFS directory on the server.
You will get "Access Denied" because ADFS is running. So you have to stop the ADFS service, copy over the .dll and then start up the service again.
You do not have to delete the custom attribute store in the wizard and reload it. When ADFS starts. it will load the latest .dll.
The gist with the code is here.
But if you want a sneak preview, the important part is:
Enjoy!
This is a good write-up.
Unfortunately, all the code is in a screen shot which sucks. Somewhat difficult to copy / paste :-).
Luckily, similar code can be found here.
Just standardise the names; one is "ToUpper"; the other is "toUpper".
However, my requirement was for getting claims from a back-end (details unimportant for the purposes of this post) where a user could have many claims of that type returned. Think of a property ID where one person owns a house but an investor owns several.
All the examples were for returning one attribute.
Essentially, you are returning a C# jagged array e.g. string[][] resultData.
My query string was :
c:[Type == "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/emailaddress"]
=> issue(store = "CAS", types = ("HouseID"), query = "House", param = c.Value);
So I assumed that the multiple claims would be in the same row i.e. one row; many columns.
The search to the back-end returned 3 houses.
I then got:
Microsoft.IdentityServer.ClaimsPolicy.Language.PolicyEvaluationException: POLICY0019: Query 'House' to attribute store 'CAS' returned an unexpected number of fields: expected '1', got '3'.
If you look at the query, you'll see there is only one query parameter which will only return one result. A query which returns 3 results would be like:
query = "House", sn, mail
So what I need is 3 rows; one column.
I found some guidance around this here.
One of the things that stumped me for a while was the fact that the array had to be dynamic because there could be any number of houses. That's why you add to a list and then cast to an array.
Another gotcha was the fact that while you can have a static rule like:
=> issue(type = "HouseID", value = "123456");
you cannot have that in a query string for the attribute store. You get:
System.ArgumentException: ID4216: The ClaimType 'HouseID' must be of format 'namespace'/'name'.
So it needs to be:
c:[Type == "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/emailaddress"]
=> issue(store = "CAS", types = ("http://claim/HouseID"), query = "House", param = c.Value);
i.e. HouseID becomes http://claim/HouseID.
Once you have compiled a new .dll file, you have to copy it over in the ADFS directory on the server.
You will get "Access Denied" because ADFS is running. So you have to stop the ADFS service, copy over the .dll and then start up the service again.
You do not have to delete the custom attribute store in the wizard and reload it. When ADFS starts. it will load the latest .dll.
The gist with the code is here.
But if you want a sneak preview, the important part is:
try
{
// Dummy values to illustrate the principle.
List<string> claimValues = new List<string>();
claimValues.Add("123456");
claimValues.Add("654321");
claimValues.Add("456123");
List<string[]> claimData = new List<string[]>();
// Each claim value is added to its own string array
foreach (string claimVal in claimValues)
{
claimData.Add(new string[1] { claimVal });
}
// The claim value string arrays are added to the string [][] that is
// returned by the Custom Attribute Store EndExecuteQuery()
string[][] resultData = claimData.ToArray();
TypedAsyncResult<string[][]> asyncResult = new TypedAsyncResult<string[]
[]>(callback, state);
asyncResult.Complete(resultData, true);
return asyncResult;
}
catch (Exception ex)
{
String innerMess = "";
if (ex.InnerException != null)
innerMess = ex.InnerException.ToString();
throw new AttributeStoreQueryExecutionException("CAS exception : " +
ex.Message + " " + innerMess);
}
Enjoy!
Tuesday, March 21, 2017
NLOG : Logging for both web site and web API
I have a standard .NET 4.5 MVC web site with web API.
I added NLOG to the web API via the NuGet packages and log away quite happily.
Part of the NuGet are the files:
So I copied the two NLog files from the web API project and pasted them into the web site project.
I changed the logging file name. So I have web-site.log and web-api.log.
Works fine.
Enjoy!
I added NLOG to the web API via the NuGet packages and log away quite happily.
Part of the NuGet are the files:
- NLog.config
- NLog.xsd
So I copied the two NLog files from the web API project and pasted them into the web site project.
I changed the logging file name. So I have web-site.log and web-api.log.
Works fine.
Enjoy!
Monday, March 20, 2017
ADFS : WIF10201: No valid key mapping found
The error is:
WIF10201: No valid key mapping found for securityToken: 'System.IdentityModel.Tokens.X509SecurityToken' and issuer: 'http://MY-ADFS/adfs/services/trust'.
I have a simple WIF application circa VS 2012 that I use to display claims and ported it over to use on ADFS 4.0.
Then I got the above error.
The solution is as per Signing key rollover in Azure Active Directory.
Yes - it says AAD but the client-side code for ADFS is the same since it's all driven from the metadata.
Use the code from: "Web applications protecting resources and created with Visual Studio 2012".
When I compared the web.config changes, the error seemed to be because the server name is "MY-ADFS" (in caps) but I had written "my-adfs" (no caps) in the web.config.
The thumbprint was also in caps. (Although I've never had an issue with that).
It gives you a nice comment:
"Element below commented by: ValidatingIssuerNameRegistry.WriteToConfg on: '20/03/2017 1:00:16 a.m. (UTC)'. Differences were found in the Metadata from: ..."
Enjoy!
WIF10201: No valid key mapping found for securityToken: 'System.IdentityModel.Tokens.X509SecurityToken' and issuer: 'http://MY-ADFS/adfs/services/trust'.
I have a simple WIF application circa VS 2012 that I use to display claims and ported it over to use on ADFS 4.0.
Then I got the above error.
The solution is as per Signing key rollover in Azure Active Directory.
Yes - it says AAD but the client-side code for ADFS is the same since it's all driven from the metadata.
Use the code from: "Web applications protecting resources and created with Visual Studio 2012".
When I compared the web.config changes, the error seemed to be because the server name is "MY-ADFS" (in caps) but I had written "my-adfs" (no caps) in the web.config.
The thumbprint was also in caps. (Although I've never had an issue with that).
It gives you a nice comment:
"Element below commented by: ValidatingIssuerNameRegistry.WriteToConfg on: '20/03/2017 1:00:16 a.m. (UTC)'. Differences were found in the Metadata from: ..."
Enjoy!
AAD : Beware the difference in the V1.0 and V2.0 endpoints
The V2.0 endpoint is the endpoint that allows you to sign in with the converged Microsoft and Azure Active Directory accounts.
From an OpenID / OAuth perspective, the discovery documents can be found at:
V1.0:
https://login.microsoftonline.com/common/.well-known/openid-configuration
V2.0:
https://login.microsoftonline.com/common/v2.0/.well-known/openid-configuration
Notice the extra "V2.0".
If we look at the keys, we see for V1.0:
https://login.microsoftonline.com/common/discovery/keys
And for V2.0 :
https://login.microsoftonline.com/common/discovery/v2.0/keys
You will notice that V2.0 has four signing keys while v1.0 only has two.
V2.0 is the endpoint used by B2C.
Don't assume that tokens signed by Azure AD (V1.0) are also acceptable for B2C (V2.0) and vice versa.
This will be true when AAD and B2C are merged at some point in the future but right now it's a gotcha!
Enjoy!
From an OpenID / OAuth perspective, the discovery documents can be found at:
V1.0:
https://login.microsoftonline.com/common/.well-known/openid-configuration
V2.0:
https://login.microsoftonline.com/common/v2.0/.well-known/openid-configuration
Notice the extra "V2.0".
If we look at the keys, we see for V1.0:
https://login.microsoftonline.com/common/discovery/keys
And for V2.0 :
https://login.microsoftonline.com/common/discovery/v2.0/keys
You will notice that V2.0 has four signing keys while v1.0 only has two.
V2.0 is the endpoint used by B2C.
Don't assume that tokens signed by Azure AD (V1.0) are also acceptable for B2C (V2.0) and vice versa.
This will be true when AAD and B2C are merged at some point in the future but right now it's a gotcha!
Enjoy!
Wednesday, March 15, 2017
ADFS : Using the client_credentials flow with ADFS 4.0 returns 401
Been battling with this for ages. I'm using ADFS 4.0 on Server 2016.
My web site uses OpenID Connect and that uses the OWIN authorisation code grant.
When I use the authorisation code grant, I can get the code, exchange the code for an access token and then call a web API with that access token no problem.
But that flow requires a user to authenticate and for some of my use cases there is no user. An example would be a forgotten password flow where the user cannot authenticate. To do that, I use the client_credentials flow.
Getting this to work was a non-trivial task since the documentation is (shall we say) sub optimal.
I also needed the ADAL Nuget package. Interesting that you can mix OWIN and ADAL
So under "Application Groups", I have the "Server application accessing a web API" scenario.
The server application has a client ID and the secret key.
So assume the general API URL is:
https://my-pc/WebService
And the API's inside this are like:
https://my-pc/WebService/api/my-api/
The web API in the wizard has the RP identifier:
https://my-pc/WebService
(This matches the ValidAudience below).
Access control policy is:
Permit everyone
I have one custom claim rule:
c:[] => issue(claim = c);
Now there is no way to pass these claims back in the JWT but I added this during my troubleshooting.
Client permissions is set to:
"All clients"
with scope of:
openid and user_impersonation.
The web API controller is decorated with [Authorize].
The code is:
To call a second API, it's the same code with one change:
HttpRequestMessage request = new HttpRequestMessage(HttpMethod.Post, "https://my-pc/WebService/api/my-api1/");
I also has to change to code for the web API App_Start, Startup.Auth.cs.
Enjoy!
My web site uses OpenID Connect and that uses the OWIN authorisation code grant.
When I use the authorisation code grant, I can get the code, exchange the code for an access token and then call a web API with that access token no problem.
But that flow requires a user to authenticate and for some of my use cases there is no user. An example would be a forgotten password flow where the user cannot authenticate. To do that, I use the client_credentials flow.
Getting this to work was a non-trivial task since the documentation is (shall we say) sub optimal.
I also needed the ADAL Nuget package. Interesting that you can mix OWIN and ADAL
So under "Application Groups", I have the "Server application accessing a web API" scenario.
The server application has a client ID and the secret key.
So assume the general API URL is:
https://my-pc/WebService
And the API's inside this are like:
https://my-pc/WebService/api/my-api/
The web API in the wizard has the RP identifier:
https://my-pc/WebService
(This matches the ValidAudience below).
Access control policy is:
Permit everyone
I have one custom claim rule:
c:[] => issue(claim = c);
Now there is no way to pass these claims back in the JWT but I added this during my troubleshooting.
Client permissions is set to:
"All clients"
with scope of:
openid and user_impersonation.
The web API controller is decorated with [Authorize].
The code is:
using Microsoft.IdentityModel.Clients.ActiveDirectory;
ClientCredential clientCredential = new ClientCredential(clientId, secretKey);
AuthenticationContext ac = new AuthenticationContext("https://my-adfs/adfs/", false);
AuthenticationResult result = await ac.AcquireTokenAsync("https://my-pc/WebService",
clientCredential);
HttpRequestMessage request = new HttpRequestMessage(HttpMethod.Post,
"https://my-pc/WebService/api/my-api/");
request.Headers.Authorization = new AuthenticationHeaderValue("Bearer", result.AccessToken);
HttpContent content = new FormUrlEncodedContent(new[] { new KeyValuePair<string,
string>("foo", "blah"), new KeyValuePair<string, string>("foo1", "blah1") });
request.Content = content;
HttpResponseMessage response = await client.SendAsync(request);
if (response.IsSuccessStatusCode)
// etc
To call a second API, it's the same code with one change:
HttpRequestMessage request = new HttpRequestMessage(HttpMethod.Post, "https://my-pc/WebService/api/my-api1/");
I also has to change to code for the web API App_Start, Startup.Auth.cs.
app.UseActiveDirectoryFederationServicesBearerAuthentication(
new ActiveDirectoryFederationServicesBearerAuthenticationOptions
{
TokenValidationParameters = new TokenValidationParameters()
{
SaveSigninToken = true,
ValidAudience = "https://my-pc/WebService"
},
MetadataEndpoint = "https://my-adfs/FederationMetadata/2007-06/FederationMetadata.xml"
});
Enjoy!
Monday, March 06, 2017
ADFS : Health Check
This is a question I've been asked a number of times.
Usually, you just ping the metadata endpoint or the IDPInitiatedSignOn endpoint.
Then I found: AD FS Diagnostics Module.
"The AD FS Diagnostics Module contains commandlets to gather configuration information of an AD FS server, as well as commandlets to perform health checks to detect configuration issues based on common root causes identified during support engagements such as duplicate SPN, cert".
There are some other useful links on the LHS.
On my list to try out :-)
Enjoy!
Usually, you just ping the metadata endpoint or the IDPInitiatedSignOn endpoint.
Then I found: AD FS Diagnostics Module.
"The AD FS Diagnostics Module contains commandlets to gather configuration information of an AD FS server, as well as commandlets to perform health checks to detect configuration issues based on common root causes identified during support engagements such as duplicate SPN, cert".
There are some other useful links on the LHS.
On my list to try out :-)
Enjoy!
Subscribe to:
Posts (Atom)