I've been assigned to find a way of implementing SSO in our products. We have several Winform applications and one asp.net 4.0 web app (not MVC).
All the products are built using .Net 4.0, the web app is ASP.NET 4.0.
Some of the Winforms are commmunicating with our API via web services (asmx) and some uses our API directly. The web app is using the same API as well. We offer a set of web services (asmx) that uses the same API to external clients.
Currently we have our own authentication implementation (user, password, roles) in our systems and we would like to replace that with SSO. Or can these two authentication regimes co-exist somehow? The Winforms are used in intranets and the web app is used both in intranets and we also hosts the web apps for clients (accesible from the Internet).
The users are created in our system, but at the same time we import users from Active Directory using our own tool. Active Directory is really the primary user source.
I have read about Windows Identity Foundation and I wonder if I can use that to implement SSO. But what I don't understand is how to use WIF in the winform applications when they use the API directly.
What I would like to achieve is to remove all user administration from our system and use Active Directory as the user source. I guess that means using ADFS 2.0 to create claims, etc.
I can use .Net Framework 4.5 in this implementation (I know that WIF is now a first class citizen in .Net Framework 4.5).
Do you have any advices how to do this? Is WIF the best alternative to achieve SSO across winforms applications and web apps?
There is a way to get the WIF authentication cookie from within the WinForms application.
To do it, you just host the WebBrowser control and point it to the login page of your web application. Assuming the web application is federated with the ADFS2, the web browser control will automatically follow the flow - it will redirect to ADFS and stop there to show the prompt for user credentials (ADFS2 in Forms Authentication mode) or just authenticate using NTLM/Kerberos (ADFS2 in Windows authentication mode). Then the web browser will redirect back to your application.
This is where you hook your code. You just add a handler to the web browser's navigation event and you check when it comes back to your application AFTER ADFS2.0 authenticates the user. You can then call the InternetGetCookie method in the WinForms app to get all the authentication cookies issued by your application and you can close the window which hosts the web browser.
At this point, you have all authentication cookies issued by WIF (the SessionAuthenticationModule) for your application. You can now call your application web services and inject cookies into http calls. The web server will correctly recognize users as authenticated which means that all you have to do is to add proper authorization to your web services (the PrincipalPermission on your web methods should do).
An alternative approach would be to expose WCF services from your web application and guard them with WS-Federation active authentication. The downside of this approach is (in my opinion) that if your identity provider (ADFS) is further federated with yet another identity provider which DOES NOT necessarily implement WS-Trust/WS-Federation then the active authentication will probably fail (because the other identity provider does not implement it) while the passive scenario will still work (a bunch of redirects will sooner ot later end with a page which requires user to provide the credentials but the flow of authentication protocols between consecutive identity providers does not matter).
Related
Is it possible to mix authentication types in a C# ASP.NET MVC web application hosted in Azure?
I have an ASP.NET MVC application written in C# that uses ASP.NET Identity as its authentication system.
A customer has asked if they can sign into the application using their Azure Active Directory (SSO using openconnect id).
I can recreate the application using openconnectid and assign their tenant as the AAD directory but users from my company can not log into the application because we do not exist in the customers Azure Active Directory.
We need to be able to log into the application because we perform data entry tasks for them using the web app.
Has anyone come across a similar issue?
Regards,
Graham
You may need to show login screen with 2 possible options. You redirect the user to their respective identity provider , they get authenticated and bring back the access_token/Id_token to access application resources.
Very much like different OpenID connect providers in the same application(FB, Google, Microsoft etc) and regardless of which provider user choose to get authenticated the token is same to access resources.
I have three asp.net applications. Only one of them has a forms authentication login. I redirect anonymous users to that one login page for all three applications to login. Once they log in, they automatically redirected back to the application and page they were attempting to access.
I enabled this functionality by setting the same MachineKey in all three applications.
Is there a way to do this for ADFS 3.0 WIF authentication as well? It doesn't seem to work the same in my testing. When I log into the application that is wired up to ADFS, I still can't access the other two.
WIF and ADFS don't work the same way as traditional forms authentication. These technologies rely on issuing access tokens, and require that dependent applications (also known as Relying Parties, or RPs) configure a trust relationship with the token provider (AKA Identity Provider, or IP). You can't share the cookie with MachineKey between apps that have not directly authenticated with an IP, and to be quite honest you don't want to.
The typical web scenario (also known as Passive Federation) is to have a separate application that functions as a Security Token Service (STS). This application houses the Login.aspx page and is protected with Forms or Windows Authentication like you would find in a classic ASP.NET scenario. When you attempt to access a web application that requires authentication, it needs to be set up to redirect you to the STS website, rather than handling it by itself. Once you log into the central STS, it will issue you a token that you then provide to applications to gain access. If you use WIF properly, this is all handled behind the scenes and is just a matter of configuration.
Each of your three web applications should be configured with a trust relationship to your IP. You said that you have a web application wired up to ADFS already, if that's via the proper trust relationship, then you should simply have to replicate that set up to your other 2 applications.
I am trying to do something like this:
I have a MVC4 Web App and a Web-API service (hosted on two separate roles in azure)
Another role runs CustomSTS1.
The MVC Web App trusts the CustomSTS1
Now the customer logs into the site he is redirected to the STS login page.
Once logged in, he is redirected back to the MVC Web Site.
From this web site, the customer performs actions, which in turn invoke the web-API Service.
I have the SAML token in the web app, which I pass to the WebAPI service.
Now when I try to validate the SAML token at the Web API side, I get a
Message=ID1032: At least one 'audienceUri' must be specified in the SamlSecurityTokenRequirement when the AudienceUriMode is set to 'Always' or 'BearerKeyOnly'. Either add the valid URI values to the AudienceUris property of SamlSecurityTokenRequirement, or turn off checking by specifying an AudienceUriMode of 'Never' on the SamlSecurityTokenRequirement.
This is without the Web API service trusting the CustomSTS1
Once I setup the trust,
I am always given a HTTP 401: UNAUTHORIZED, whenever I try to make a HTTP Get request to the WEB API Service.
Now, My Question is, (I know that my current approach is definitely wrong)
How do I setup the Trust relationship with the CustomSTS1, such that the WebAPI service is able to do an ActAS on behalf of the user logged into the MVC site?
OR
Is this architecture wrong?
And is there another way to achieve this?
That approach is wrong conceptually. The MVC application should negotiate a new token for the Web API in the STS using ActAs. That's how it traditionally works for SOAP Services. However, Web APIs are moving away from SAML as it is a complex format that relies on different WS-* specs. OAuth 2.0 is becoming the standard in that area if you want to support SSO at that level.
Another approach is to establish an implicit trust between the MVC app and the Web API, so all the calls to the Web API from the MVC app are done through a more standard Http auth mechanism like Basic Auth using an specific set of credentials that only the MVC app knows. The info about the logged user in the MVC app is passed as additional information.
Regards,
Pablo.
What are the authentication options for having a ASP.NET web application communicating with a WCF service?
The scenario:
User enters their username and password in an ASP.NET form.
ASP.NET needs to pass this to WCF to authenticate the user.
If authenticated, the user can perform actions on the website. Each action would require sending data to different WCF operations. WCF needs to know who the user is on each call.
The easiest solution would be to store the username/password in the ASP.NET session state. However, this is insecure because the password is stored in memory on the server.
I would rather not use a certificate to authenticate the ASP.NET "client" to the service because there's a possibility that this WCF could be consumed by another client in addition to ASP.NET.
The best suggestion I've seen so far is to use Windows Identity Foundation (WIF). It appears that this requires an STS. According to MSDN, Microsoft does not seem to recommend setting up an STS through Visual Studio. There's no guarantee that an STS would be available in the deployment environment as some environments may use Active Directory and other environments may have a custom user store. Is it possible to setup a custom STS to authenticate against a custom user store? I'm having trouble finding documentation on doing this.
Are there any other options besides using WIF? What about a custom WCF authentication service that returns a token that can be used for authenticating against a primary WCF service?
The standard way of doing this is by using WIF with Microsoft's STS viz. Active Directory Federation Services v2.0 (ADFS).
There are a number of custom STS available e.g. Identity Server. This use a SQL DB as an attribute store. It's open source so could be adapted to whatever you require.
You can create your own custom attribute store: AD FS 2.0 Attribute Store Overview.
TechNet WIF / WCF: WIF and WCF.
Here is my scenario -
1/ I have an ASP.NET MVC application running on my server, it uses Windows Authentication.
2/ There is different web application (written in Java) somewhere else that also uses Windows Authentication.
In the Controller of my MVC application I need to grab some information from this other Web app. How can I connect to the "foreign" application using the credentials of the user that is accessing my Controller?
Any help appreciated.
Impersonation doesn't leave the ASPNET process. This means that you cannot delegate the credentials and access a remote resource using them. You could either swap back to Forms authentication which uses cookies or you will need Kerberos.