Hello all. You may recall from older posts on this blog (2012) that we’ve played around in the past with Home Realm Discovery (HRD) in AD FS.
First with IWA and forms logon here and then a little bit more for good measure. Since then it’s been a little quiet on the HRD fence and not a subject delved into much since. That is, until an issue a colleague and I ran into an issue recently… confronted with the following AD FS R2 access conundrum:
“ How can we route (automatically) external logon requests to a third-party claims provider, continue to route internal logon requests to the native AD claims provider (i.e. maintain transparent SSO), eliminating any Home Realm Discovery (HRD) for our users to have to walk through and select?”
Or… to put it more simply. How can one automatically route external route requests to claims provider A, continue to route internal requests to (AD) claims provider B, whilst altogether avoiding Home Realm Discovery (HRD) for all users?
We’ll discuss the exact use case in a moment. This is a RP-initiated work-flow. No IDP-initiated flow or smart links are allowed.
First, a bit of background. Let’s head back to the first look posts on AD FS 2012 R2 in 2013/2014. In there I wrote:
“In AD FS R2, HRD customization options through PowerShell now allow us to determine how the UI is presented to the end-user during logon, based on:
1. target suffix resolution – providing suffix routing on a per claims provider basis
2. limiting the claims providers visible for the relying party concerned”
With the move to kernel-mode in R2, customization previously possible under IIS in AD FS 2.0 was no longer applicable. To redress this issue and to provide greater flexibility, the above customization options were provided. Suffix routing enables AD FS 2012 R2 to select an appropriate claims provider based on the domain suffix credentials provided by that user. Broadly speaking, this is just how O365/Azure AD works in attempting to discover the Federated home realm v Microsoft realm of the connecting user. Similarly, for on-premise AD FS, suffix routing enables us to refine our logon decision-making based on the home realm of the connecting user. Here’s an example:
In AD FS 2012 R2 we can configure this within Powershell:
Set-AdfsClaimsProviderTrust –TargetName “Cranky Nuts IdP” –OrganizationalAccountSuffix @(“crankynuts.nl”)
Set-AdfsClaimsProviderTrust –TargetName “Wobbly Wheels IdP” –OrganizationalAccountSuffix @(“wobblywheels.be”)
Wonderful! Using this approach can be a little hit and miss though. It’s making certain assumptions about the level at which we’re prepared to divert authentication on, i.e. the domain level. That’s not always the case.
We could, alternatively, opt for the other method: targeting claims provider selection at the Relying Party Trust level. Again, this is useful functionality. Here’s an example.
Set-AdfsRelyingPartyTrust -TargetName “BodgeIT Inventory ” -ClaimsProviderName @(“crankynuts.nl”)
This makes the “BodgeIT Inventory” Relying Party/Web Application, available to users of the Cranky Nuts organization, bypassing the local AD claims provider. Should we need to extend access to additional claims providers, such as our local Active Directory, then we can add that issuer to the mix. The downside of this option, is that we’re forced into making a home-realm discovery selection during logon.
The Suffix and RP-specific claims provider options, therefore, do have some limitations. Indeed, for the scenario that we were faced with, neither option fit. In our situation, two claims providers existed for a single organization, with each CP containing the same set of users. The use case? Two-factor authentication. The upstream claims provider (CP) provides two-factor authentication (2FA) for users connecting from the outside , while the incumbent local AD claims provider covers local Windows logon.
Of course, two claims providers means Home Realm Discovery (HRD) and HRD can be confusing. We can remedy this for internal users, ensuring they don’t get the HRD prompt by defaulting to the use of the local AD claims provider via Powershell:
Set-ADFSProperties –IntranetUseLocalClaimsProvider $True
This doesn’t solve the problem for access from the outside. Moreover, we expressly wish to send all connecting users to our 2FA claims provider, rather than to the in-build AD claims provider…. We could use auxiliary authentication provided by the MFA Adapter in AD FS 2012 R2, thereby eliminating the use of the upstream claims provider. However, that might not be feasible as:
Even where the latter is available, the (security) policy prerogative might be that use of 1) is not allowed, thereby overriding (2).
<PLUG>To demonstrate issue, cue favorite authentication platform PointSharp (</PLUG>
PointSharp possess both a Security Token Service (STS) that can stip atop AD FS as a 2FA claims provider role and also have an MFA adapter for AD FS 2012 R2 / 2016. Since use of the MFA adapter has been prohibited by the customer according to the rules above (not wanting to lead with AD password as a primary authentication credential), the STS as a Claims Provider was selected instead.
You’ll recall we turned off HRD for internal users by setting the IntranetUseLocalClaimsProvider setting to TRUE. With the PointSharp STS configured as a claims provider, external users will get continue to get the HRD screen as AD FS cannot determine where the user lives without the appropriate “hint”. In AD FS 2.0 it was possible to workaround this by customizing the AD FS 2.0 proxy configuration, modifying the Page_Init section of the homerealmdiscovery.aspx.cs file. We could then remove Active Directory as a claims provider (0)
PassiveIdentityProvidersDropDownList.Items.RemoveAt(0);
Once that was done, default realm selection would fall to our other claims provider (STS)
SelectHomeRealm ( PassiveIdentityProvidersDropDownList.SelectedItem.Value );
As you are no doubt aware, IIS has been supplanted in AD FS 2012 R2 with everything running in kernel-mode. Thankfully, all is not lost though. Some of the functionality previously possible under IIS has moved to Javascript and customization of the onload.js script is covered in a couple of scenarios here..
Variables previously living in .NET code-behind are visible in the View Source of the web page of the HRD screens. Options for both claims providers are visible.
….HRD.selection(‘http://adfs.mydomain.com/adfs/services/trust’);” onclick=”HRD.selection(‘http://adfs.mydomain.com/adfs/services/trust’); …….
….HRD.selection(‘https://sts.mydomain.com/login/’);” onclick=”HRD.selection(‘https://sts.mydomain.com/login/’); …….
HRD.selection is the main feature of interest. There’s an action on clicking the realm manually that specifies the HRD we require. We can automate this by moving the selection to the onload.js script, thereby automating selection of the PointSharp Claims Provider. First, we need to customize our AD FS setup.
We create a new theme called PointSharp based on the default AD FS theme
New-AdfsWebTheme –Name PointSharp –SourceName default
Export the default theme to allow our customization
Export-AdfsWebTheme –Name default –DirectoryPath c:\scripts
Copy the default onload.js to the C:\Scripts folder and edit it. At the end of the script, we simply add the reference to the PointSharp claims provider using the following syntax (replace with your own CP) :
HRD.selection(‘https://sts.mydomain.com/login/’);
We then import the new PointSharp theme including the modified onload.js.
Set-AdfsWebTheme -TargetName PointSharp -AdditionalFileResource @{Uri=’/adfs/portal/script/onload.js’;path=”c:\scripts\onload.js”}
Activate the new theme.
Set-AdfsWebConfig -ActiveThemeName PointSharp
External users will be now routed to the PointSharp claims provider for login, whilst AD logon for internal users is still preserved. Please note that this is a GLOBAL change affecting all users of the AD FS farm and test extensively as it might not be your cup of tea…
I recently added my O365 tenant, for testing purposes, to a AD FS in Windows Server 2016 TP4 and noticed something rather unusual. Via the AD FS Management snap-in it was not possible to assign an access-control policy in AD FS to my Office365 Relying Party (RP). Looking at my RP Trusts, I could see the Access Control Policy section was blank.
Right-clicking over a Relying Party in AD FS 2016 TP4 reveals an additional menu option for editing Access Control Policies:
Doing this on the O365 RP, only the Issuance Authorization Rules pipeline is visible, nothing else.
If we compare that with a normal RP.
Right-clicking and selecting access control policy, the administrator is presented with the option of choosing an appropriate policy for that RP: in our case, the desire to use multi-factor authentication.
Why this does not appear on the O365 pipeline, I can only speculate. It might be partially explained by the fact that creating the Office 365 Identity Platform relying party is normally performed via PowerShell and Microsoft wanted to keep this procedure ubiquitous across all versions of AD FS since v2.0. For example, the following command will create the Office 365 Identity Platform RP should it not exist.
Update-MsolFederatedDomain -SupportMultipleDomain -DomainName mydomain.com
Alternatively, following the principle of Occam’s Razor, it could be Microsoft have simply not got round to updating their code to detect AD FS 2016 during O365 Relying Party creation 
Either way, from previous experiences of using MFA under Office 365 via AD FS 2012 R2, we do know it IS possible to use MFA with O365, so getting it working with AD FS 2016 just requires a little more effort. ..
Looking at our O365 RP in PowerShell (Get-ADFSRelyingPartyTrust) , we see no access policies configured. Under the O365 Relying Party it’s blank.
Compare this to an RP that does have an access control policy configured:
Since the UI doesn’t allow enabling MFA in an access policy for our O365 RP, playing around with PowerShell reveals that it is possible using the Set-ADFSRelyingPartyTrust cmdlet .
Set-AdfsRelyingPartyTrust -AccessControlPolicyName ‘Permit everyone and require MFA’ -targetidentifier https://login.microsoftonline.com/extSTS.srf
Check the O365 relying party (Get-ADFSRelyingPartyTrust) that an Access Control Policy has been added.
In the GUI, we then see the applied policy (Permit Everyone and require MFA) appearing.
Connect to Office 365 and we’re redirected to our AD FS instance. Enter the AD credentials for the user and then MFA kicks in:
As ever, play and test at your own peril
In the next few posts, I wanted to take a look at the changes to be found in Windows Server 2012 R2 with respect to Active Directory Federation Services (AD FS). At TechEd Europe, I was fortunate enough to chat with some of the folks from the Active Directory team about the new enhancements and to cover them here in a little more detail. As you may have read, there are a significant number of changes in the R2 version and I’ll spread coverage of this over a number of posts.
Part 1
– Architecture Changes
– Workplace Join / Bring Your Own Device (BYOD)
o Windows 8.1
o iOS Devices
– Web Application Proxy
– Extranet soft account lockout policies
– Lost Device Protection
Part 2
– UI Changes
– Access Scenarios
– Authentication changes
o Multi-Factor Authentication
o Context-based Authentication
o Claims/Non-claims aware applications
o Policy-based Evaluation
Part 3
– OATH2 support
– Work Folders
– Better event handling / reporting
– Server Core
– Other stuff undiscovered J
The use of IIS with AD FS in Windows Server 2012 R2 has been eschewed in favour of a move to kernel-mode (HTTP.SYS). The motive, highlighted in discussions at TechEd, is to improve performance, provide greater sign-in customization options and to assuage concerns for co-locating AD FS and AD DS on the same server (IIS on domain controllers has been a long-standing security no-no). As the use of federation services goes more mainstream in everyday use with Windows 8.1, this shift is understandable and an important design consideration. With the new kernel-mode approach, support for running under server core also appears as an option in the new release.
From a basic architecture standpoint and overview, the AD FS proxy has been supplanted by a role known as the Web Application Proxy, servicing connections for external clients. The user interface (UI) through the migration to kernel mode is also significantly changed. Authentication undergoes a radical overhaul with a Multi-Factor Authentication (MFA) Adapter available for plugging into Windows Azure Active Authentication and third-party MFA providers. This is also seen in more nuanced behaviour with respect to authentication within the product, reflected in greater flexibility in access control decisions.
With AD FS now built directly built on top of HTTP.SYS, a lot of changes are abstracted from the user through the new MMC UI and also PowerShell. Nonetheless, it’s worthwhile familiarizing ourselves with kernel mode elements, as they serve a useful role in basic service troubleshooting/configuration . The NETSH HTTP command can be used to query and configure http.sys.
The netsh http show urlacl command can be used to list URL reservations that AD FS makes within http.sys. Here we can see the /adfs/ path reserved for use.
Reserved URL : https://+:443/adfs/
User: NT SERVICEadfssrv
Listen: Yes
Delegate: Yes
SDDL: D:(A;;GA;;;S-1-5-80-2965554544299-213434830-363436364-117610243-975697593)
And there’s this new guy, the Device Enrolment server, whose role becomes more apparent should we wish to make use of new Windows 8.1/iOS client (mobile) integration features.
Reserved URL : https://+:443/EnrollmentServer/
User: NT SERVICEdrs
Listen: Yes
Delegate: Yes
SDDL: D:(A;;GA;;;S-1-5-80-1321940109-3370001082-3650459431-215109509-2472514016)
SSL bindings can be reviewed using the netsh http show sslcert command. The AD FS server in this demo setup I’ve created is sts.adfs2.net.
PS C:Windowssystem32> netsh http show sslcert
SSL Certificate bindings:
————————-
Hostname:port : sts.adfs2.net:443
Certificate Hash : 1f54c1c62b057dscffgb1aec2b2cbd0876e5c559
Application ID : {5d89a20c-beab-4389-9447-324788eb944a}
Certificate Store Name : MY
Verify Client Certificate Revocation : Enabled
Verify Revocation Using Cached Client Certificate Only : Disabled
Usage Check : Enabled
Revocation Freshness Time : 0
URL Retrieval Timeout : 0
Ctl Identifier : (null)
Ctl Store Name : AdfsTrustedDevices
DS Mapper Usage : Disabled
Negotiate Client Certificate : Disabled
Hostname:port : localhost:443
Certificate Hash : 1f52c0d62b0570c6a26c7fec2b2cbd0876e5bc59
Application ID : {5d89a20c-beab-4389-9447-324788eb944a}
Certificate Store Name : MY
Verify Client Certificate Revocation : Enabled
Verify Revocation Using Cached Client Certificate Only : Disabled
Usage Check : Enabled
Revocation Freshness Time : 0
URL Retrieval Timeout : 0
Ctl Identifier : (null)
Ctl Store Name : AdfsTrustedDevices
DS Mapper Usage : Disabled
Negotiate Client Certificate : Disabled
Hostname:port : sts.adfs2.net:49443
Certificate Hash : 2f5c41c62b0570c6a26c7fec21d2d0876e5c559
Application ID : {5d89a20c-beab-4389-9447-324788eb944a}
Certificate Store Name : MY
Verify Client Certificate Revocation : Enabled
Verify Revocation Using Cached Client Certificate Only : Disabled
Usage Check : Enabled
Revocation Freshness Time : 0
URL Retrieval Timeout : 0
Ctl Identifier : (null)
Ctl Store Name : (null)
DS Mapper Usage : Disabled
Negotiate Client Certificate : Enabled
Location-wise, the AD FS application files themselves are no longer held under C:Program FilesActive Directory or C:Program Files (x86). Instead, they’ve moved to C:WindowsADFS. For clarity, this was actually a change instigated first in Windows Server 2012 with the Active Directory Federation Services (AD FS) 2.1 role. In this folder is the Microsoft.IdentityServer.Servicehost.exe.config file, where, as admins, we’ll be spending more time in the future in order to activate debug functions. From this file all trace options for various services and endpoints can be enabled. In the same folder is a configuration file for the new Device Registration service (DRS), responsible for activation and enrolment of controlled devices and represented by a new (Win8/IOS *) schema class in Active Directory Domain Services (AD DS). The file in question is called Microsoft.DeviceRegistration.ServiceHost.exe.config.
Support for the new Device class requires a schema change to Active Directory. For those upgrading an existing Windows setup, the appropriate files can be found on the R2 installation CD under D:SupportADPrep. From what I’ve seen/tested thus far, to support the new release, you’ll need at least one Windows Server 2012 domain controller, preferably two in any serious deployment scenario. This requirement stems from the use of Group Managed Service Accounts (GMSA) that are generated and maintained by the Key Distribution Service (KDS) on 2012 domain controllers. The new version of AD FS makes use of these GMSA accounts, defined during AD FS installation, that are then shared amongst connecting AD FS hosts. I suggest reading the following backgrounder and bear in mind that the AD FS Windows Server 2012 preview labs incorporate a workaround for testing purposes, in activating the root key, that is not recommended for production environments.
Update 29/10 – SamD from the AD product team added that “gMSA is not required to be the service account that ADFS runs on. It is an additional optimization that is available to customers if they have Win2012 domain controllers available.” The traditional service account option is available during installation.
Moving on, let’s take a look at the “broader” access audience that the new version emphasises. This can be immediately seen by viewing the claims descriptions list surfaced on a new AD FS installation.
There are around 40 new claims descriptions available in the AD FS Windows Server 2012 R2 release. As we’ll see, use of these new claims types allow us to make more refined assessments concerning access to web applications and resources.
Through the new Workplace Join feature within R2, AD FS becomes a focal point for mobile access in the enterprise and an integral component in the Microsoft Bring Your Own Device (BYOD) vision. Workplace Join allows hitherto unmanaged/untrusted operating systems such as Windows RT/Windows 8 and IOS to be moved into a more controlled access context, by allowing their registration and affiliation with Active Directory. Devices will register with Active Directory through a Device Registration Service (DRS) and subsequently use an X509 certificate bound to the user context(s) on that machine for device authentication. In a default configuration, users will login via AD FS to initiate the join process using their AD credentials. To further secure this process, additional factors can be also used with Windows Azure Active Authentication (PhoneFactor) or a third-party authentication provider exposed through the new AD FS MFA SDK.
Devices that are workplace-joined emit additional claims during the logon process. These include:
Certificate support in claims handling has also been enhanced.
Windows 8.1
In order to provide a comparison between old and new with Workplace Join, I began by looking at what claims (and any new ones) are processed from a vanilla Windows 8.1 Pro domain-joined machine, using a simple WS-Federation relying party to validate claims emitted the client and AD FS components.
Firstly via the internal network and the AD FS farm using Internet Explorer. Here the browser uses Integration Windows Authentication/Negotiate and the user silently accesses to the WIF relying party via Kerberos. The usual configuration caveats apply here: the URL of the AD FS and RP instance are in the Local Intranet Zone of IE.
To demonstrate a new change, I installed Mozilla Firefox and repeated the logon process. Instead of the Integrated Windows Authentication (IWA)/Negotiate process, the user is presented with a forms sign-in page. This represents a departure from the user experience and behaviour in AD FS 2.0. With the latter, authentication would be downgraded to NTLM, because IWA was assumed in the farm configuration and the browser needed to be configured to explicitly support Kerberos for seamless login. Where the latter action was not performed, the user would receive an NTLM challenge/response prompt, often causing confusion. With the new release, support for IWA is now governed through setting registering User Agent types within AD FS that are capable of supporting Negotiate. Out of the box, this is constrained to IE, meaning any other browser will revert to using forms logon when accessing resources from an internally connected client . Here we see the claims output from a Firefox login:
In internal login scenarios (IE/Firefox), we see new claims types emitted concerning location (whether the login request is sourced within the Corporate Network), an Application Identifier (corresponding to the Relying Party Identifier), the Client source IP (or translated) address, an Authentication Method Reference and a Client Request ID.
Then, via an external network through the new Web Application Proxy:
In addition to those claims types mentioned earlier is a new claims type for the client forwarded IP (x-ms-forwarded-client-ip) processed at the Web Application Proxy. The insidecorporatenetwork value is now set to false, as we’re on an outside network.
You may have observed at this point that there are no Device claims. This makes sense if we consider that their use is limited to client types that declare them, i.e. only workplace-joined clients currently make use of the Device class.
Onto the workplace join process itself. To get your test lab up and running, I recommend reading this TechNet article.
If you follow the lab guide carefully, and, in particular, with emphasis on getting the dependent infrastructure working correctly, then with a little patience and time, you’ll be up and running. I tried this with the basic contoso.com lab setup (a good starting point) and then expanded this in new setup using my own test domain. There are a few gotchas worth pointing out, hoping that you’ll avoid my growing pains …
1. Don’t use Cryptography Next Generation (CNG) algorithms when configuring your AD Certificate Services. They won’t work.
2. For simplicity in your CA configuration, I’d opt for use of HTTP Distribution endpoints, particularly if you intend to test “outside” configurations using the Web Application Proxy. Ensure the Certificate Revocation List (CRL) on the Certificate Distribution Point (CDP) and your Authority Information Awareness (AIA) URLs are setup correctly and reachable from the Win 8.1 client. If you’re using Delta CRLs and IIS as the web server for your CDP, don’t forget to allow Double Escaping on IIS in the Request Filtering section.
3. Ensure the Enterprise CA is trusted by the client and the certificate is installed in the Trusted Root Authorities section of the client. Importing the cert via the AIA endpoint is a good way of testing its availability and installing the certificate. Again, the certificate distribution point (CDP) URLs should be visible to the client.
4. Issue the AD FS certificate, complete with SAN for the Device Registration Service (DRS), before you begin your AD FS setup. The common name (CN) on the certificate should be the AD FS URL and two Subject Alternate Names (SAN) entries should contain the AD FS URL and one for the Device Registration Service (DRS) provided. For example, I’ve used an example using the MMC snap-in Certificate Request wizard, based on a copy of the default Web Server template for a common name of sts.mydomain.com for the FQDN URL of the AD FS service and also as the first Subject Alternate Name (SAN) entry. A second SAN entry is used for the DRS endpoint, enterpriseregistration.mydomain.com.
5. As posted on the Technet forums (in the current R2 preview), both the AD FS server and the Windows 8.1 client need to be configured with time-zone of UTC 0:00 or less. My setup is using Pacific Time (UTC -8:00). Don’t ask me, I just blog here J
6. Don’t forget to enable Device Authentication in Global Primary Authentication settings within the UI or via PowerShell.
7. Use the Event Log Microsoft|Workplace Join to troubleshoot!! Here’s a collage of first-hand events to illustrate it’s effectiveness in troubleshooting:
URL (enterpriseregistration.xxxx.yyyy) cannot be resolved or reached.
Can’t reach the CRL CDP of the AD CS endpoint.
Root Authority is not trusted by the client.
This message is stating a number of possible issues (generally bad):
o DRS Configuration and Activation was not completed successfully
o Issues with the SAN on the SSL certificate
o AD FS Device Authentication has not been enabled
Once you start seeing messages like the following, you’re almost there.
Nice.
8. Take particular note of any errors reported when trying to activate Device Registration Service; namely anything along the lines of:
WARNING: UPN values that are not included in the SSL certificate have been found in the enterprise. Users with these UPN suffix values will not be able to register their devices. To enable users with the corresponding UPN suffix to register their devices, provide a new SSL certificate containing the values listed below in the subject or subject alternative name.
enterpriseregistration.upn1.contoso.com
enterpriseregistration.upn2.contoso.com
In the case of (8), I’d made the mistake of not registering the appropriate certificates with Subject Alternate Names (SAN) that included the DeviceRegistration CNAME record. Simply re-issuing the AD FS service certificate afterward, setting Manage Private Keys etc. and re-activating DRS in PowerShell was not sufficient to get the configuration working.
The Workplace Join function can be accessed by first accessing the Change PC Settings option on the Windows 8 UI
In PC Settings, choose the Network option
Then select Network followed by the Workplace option:
If your configuration is working, certificates are trusted, appropriate AD FS and PKI endpoints are reachable, stars are in alignment (just joking), then clicking on the Join button leads to AD FS responding with a challenge:
Enter the Active Directory credentials for the user. In this example I’m using, the device is joining a test domain called adfs2.net. Note the AD FS URL (connecting to my R2 instance sts.adfs2.net) at the top of the page. For the auto-discovery of the AD FS Device Registration Endpoints (DRS) a CNAME (Alias) record in DNS needed to be created for the service called enterpriseregistration.adfs2.net. This record points to the host (A) record of the AD FS federation service internally. This allows the discovery process to find the DRS endpoint and in an external setting this would point to the Web Application Proxy, your own Reverse Proxy or other suitable edge device.
The relying party (RP) for the Device Registration Service is created during the DRS activation process, so there’s nothing additional required on this side.
Connecting Windows 8.1 clients will use the auto-discover function by matching the domain suffix of the user account provided during the join process against the enterprise registration CNAME record for that domain. The join process then attempts a call to the enrollment server web service. Using the adfs2.net domain as an example, the following endpoint is queried:
https://enterpriseregistration.adfs2.net/EnrollmentServer/contract?api-version=1.0
If the service can be reached successfully, the join process is initiated.
The process is now completed and the “join” associated with the Windows 8.1 user profile. I used a Microsoft Live ID account and as can be be seen from the above screenshot, a subsequent AD user called demo with a UPN of demo@adfs2.net,in doing so providing my AD credentials during the Join. Please note that the Active Directory domain I’m using is also called adfs2.net (dc=adfs2,dc=net) for convenience. In a real-world/production scenario, the DNS domain names used for Active Directory and that of the federation service itself may be different.
The user account is then issued with a self-signed X509 certificate with an Extended Key Usage (EKU) of Client Authentication. Jumping into the Certificates|User snap-in we see a certificate issued under the user context.
Back to the WIF 3.5 relying party (RP), logging on to the RP from the outside we get redirected to AD FS for logon. Upon successful logon, the following claims are shown.
With the Win 8.1 device now connected to the AD domain via a Workplace Join, we see additional claims consummated.
· IsRegisteredUser
· OS Version
· OS Type
· Identifier (Subject name of the cert)
· Display Name (corresponding to the Device Name)
· Registration ID (corresponding to an OU served up in the certificate)
The device itself is registered within Active Directory at the following location: CN=<Device ID>,CN=RegisteredDevices,DC=mydomain,DC=com.
Here’s an example with a SAML 2.0 Service Provider (SimpleSAMLphp), with a Workplace Joined Windows 8.1 client connecting to it.
Inside the corporate network, we see the following:
As tempting it is to delve further into authentication, I’ll refrain from doing so and leave this to a follow-up post. My apologies, otherwise this post will reach biblical proportions in length and we’ll be shaking hands with Santa Claus before we know it.
iOS devices
From testing, the auto-discover function using the enterpriseregistration CNAME record in DNS, described in the previous section, is limited to the workplace join process for Windows 8.1. iOS clients must directly connect to AD FS to initiate the join process. The endpoint settings on the DRS Relying Party refer to a URL of:
https://sts.adfs2.net/Enrollment Server/otaprofile/
This is the DRS Over-the-Air endpoint for non-Windows devices (currently iOS only).
I used an iPad 3 running iOS 6.1.3 for this exercise. If you plan on using self-signed certificates for this type of testing, you’ll need to use the iPhone Configuration Utility to create a profile which can be then used to install the root certificate from your Certificate Services Issuing CA (and any optional chain).
If you’re testing iOS or Windows 8.1 devices in an external setting, it’s worth mentioning that the Web Application Proxy (WAP), which I’ll cover in a moment, doesn’t provide an HTTP Reverse Proxy function. To ensure that any CRL/AIA distribution points are visible in an “outside” testing context, I elected to install IIS on the WAP, publish the CRL/AIA Certificate Distribution Points (CDP) via a UNC from the CA itself to be made available as an HTTP URL on the Web Application Proxy via IIS, making the distribution points reachable from an external perspective. Clearly, this is not something one would do automatically in a production environment, without a bit of forethought, but it works well in a demo environment. You can probably also do this with a kernel mode only approach, but I didn’t have time to test this (yet).
Once the Apple configuration file (.mobileconfig) file had been deployed onto my iPad (via e-mail), a Profile containing the Root certificate was generated and the certificate installed.
With the root certificate or chain correctly installed, going to the AD FS server URL, we should not receive any SSL errors or warnings in the browser, indicating that the chain and CRL/AIA distribution points are reachable. To test this, you can use the IdP initiated sign-on page in the default setup, e.g. https://YOURFQDN/adfs/ls/idpinitiatedsignon.aspx. From the iPad, here’s the portrait view of the page.
Playing around, I turned the iPad on its side and we get an automatically resized window. This is a nice feature in the new UI in AD FS 2012 R2 that supports dynamic adjustment and positioning of elements through CSS, resizing pages accordingly across various devices and user agents (think mobile client).
In order to kick off the Workplace Join, we point Safari to the endpoint DRS mentioned earlier:
https://sts.adfs2.net/Enrollment Server/otaprofile/
This redirects the browser to a sign-in page where we need to logon with the AD account that will be bound to the iOS device for the workplace join.
Logging on with the demo@adfs2.net account I used in the Windows 8.1 example, the Safari page remains open in the background and the foreground switches to the install profile option on the mobile device.
The install profile option and Workplace Join install option appears:
Clicking on the More Details option, we can see that the AD FS Token Signing Certificate (public key) and the Device Enrollment Encrypted Profile Service are referenced during the profile installation.
Clicking on Install Profile
Once the profile is installed we see a Certificate issued to the device, issued with a common name of MS-Organization-Access, as per the Windows 8.1 join process.
Returning to the profile screen we see the completed Workplace Join profile
NB: The Demo Auth360 profile is the imported .mobileconfig containing the root certificate from earlier.
Sitting in front of the AD FS farm is a new optional role, similar to the AD FS Proxy in AD FS 2.0, called the Web Application Proxy. This is a completely redesigned component, built to cater for federation services scenarios as well additional access scenarios beyond those seen in AD FS 2.0.
As with DirectAccess in Windows Server 2012, more roles are being moving into the mainstream product and the Web Application Proxy is a module in the Remote Access role within Windows Server 2012 R2.
Configuration of the proxy itself also moves to the Remote Access Management snap-in.
A configuration wizard is provided to connect the proxy to the back-end AD FS farm and a service account is required to register with the AD FS server(s) during installation. This service needs to be a member of the local administrators group on the AD FS farm.
Once connected to AD FS, a number of simple options are available for configuration.
The UI is at this stage is admittedly basic, but as with DirectAccess in 2012, there’s a greater emphasis on using wizards to get the job done and whatever can be done in the UI can be done (and more) via PowerShell; a PS configuration script is provided as a summary at the end of each publishing wizard rule to demonstrate this point.
As with TMG/UG we can publish/proxy a particular URL or URIs/paths of that URL expressed as separate publishing rules on the proxy , e.g.
www.mydomain.com/ (as one allow all rule) versus
www.mydomain.com/app1/ as Rule#1
www.mydomain.com/app2/ as Rule#2
www.mydomain.com/app3/ as Rule#3
An interesting under-the-covers capability is support for Server Name Indication (SNI). SNI, initially provided in Windows Server 2012, allows for multiple certificates to be bound to a single IP listener. Prior to IIS 8.0/SNI, sharing IP addresses amongst multiple websites was limited to the network endpoint and their IP:Port binding. SNI is a TLS extension that provides the hostname of the server the client is connecting to during handshaking. This allows much greater flexibility when configuring the proxy. With the move to kernel-mode, all the hard lifting is done through the UI or via PowerShell. familiarity with NETSH HTTP will also assist in any troubleshooting or ad-hoc configuration. The majority of browsers support SNI, although Windows XP in any configuration using Internet Explorer does not.
The current preview of the proxy in the R2 release provides for connections as:
1. A reverse web proxy, connecting to back-end servers via HTTPS;
2. A pre-authentication web proxy, connecting to AD FS via HTTPS to validate credentials
a. For claims aware web applications
b. For non-claims aware web applications using Kerberos
When we publish a new AD FS compatible application (pre-authentication), the proxy pulls the RP list/ configuration from the AD FS farm. Polling is done (looking at the event logs) every minute.
Using a Windows Identity Foundation (WIF) test relying party application from another test domain (psid.local), here’s an example of a publishing rule:
Add-WebApplicationProxyApplication -BackendServerUrl ‘https://rp.psid.local/app9/’ -ExternalCertificateThumbprint ’91D8014979B9CDEF9C907171F7CE9AF398E66DC6′ -ExternalUrl ‘https://rp.psid.local/app9/’ -Name ‘WIF Test Application’ -ExternalPreAuthentication ADFS -ADFSRelyingPartyName ‘WIF 3.5 Application’
This is a pre-authentication rule, meaning that AD FS process the login request, UI surfaced and validates access and authentication credentials through the proxy via a back-channel connection before access to the relying party is processed.
To complete this round of testing, I wanted to validate Workplace Join from an “outside” network via the proxy. DRS endpoints are automatically published, meaning no specific additional publishing rules needed to be created.
On the Windows 8.1 client, I removed the client from the Join agreement created earlier in order to re-attempt the join via the proxy from the external network. When we attempt a join again from the “outside” via the Web Application Proxy, clicking on Join generates the following page.
The user ID is automatically populated in the form, carried over from the Join request.
Testing from the “outside” now, when we access our WIF application RP from our Workplace Joined client (via the Web Application Proxy), we’re served up with a sign-in form.
That’s expected as we’re now an Extranet client (using MS terminology). This is confirmed by looking at the base AD FS configuration and the primary authentication provider serving up forms login for external clients.
“Extranet” users are automatically assigned to the forms authentication sign-in process, whereas Intranet users are assigned Windows Authentication, browser considerations notwithstanding. For those familiar with fiddling with Local Authentication Types in web.config on the AD FS proxy/farm in AD FS 2.0, making this available through the UI and Powershell is a boon J
Going back to our relying party application, we can see in the produced claims that the client connection is not through the corporate network and is via Web Application Proxy using the following claims:
http://schemas.microsoft.com/ws/2012/01/insidecorporatenetwork
http://schemas.microsoft.com/2012/01/devicecontext/claims/isregistereduser
The client is outside the corporate network and the user is now registered. As a simple test, if we only want to allow Registered Users access to our RP claims web application, we could do this through an Authorization Rule that states that only Registered Users are permitted access:
c:[Type == “http://schemas.microsoft.com/2012/01/devicecontext/claims/isregistereduser”, Value =~ “^(?i)true$”] => issue(Type = “http://schemas.microsoft.com/authorization/claims/permit”, Value = “PermitUsersWithClaim”);
Logging on with a non-workplace joined client from outside the corporate network, we are denied access.
Error pages are customizable and that’s something we’ll cover in Part 3.
As I touched on earlier, we can use SSL bridging scenarios incorporating wildcard certificates on the front-end (proxy) and named certificates (on the back-end) in publishing scenarios. The pre-authentication parts allows integration with claims and non-claims aware (Kerberos) applications. Applications such as Exchange, which are not claims-aware or non-SAML kerberos claims SharePoint web applications can be configured via the Web Application Proxy. From testing rich client applications such as ActiveSync and Outlook, which use Basic/NTLM application, these are not currently supported on the Web Application Proxy either in a pre-authentication or pass-through capacity. The proxy defaults to SNI and this is not supported by some mail clients. We’ll cover this and other authentication scenarios in Part 2.
Just to wrap up, here are some observations from testing:
– The proxy can translate host names in URLs but not path names. Make sure that the published path matches that of the application;
– There’s no Edit function in the UI once you’ve created a publishing rule;
– The Web Application Proxy is currently HTTPS only, so no HTTP publishing. Hopefully this will be corrected in the near future as scenarios such as CRL/CDP publishing, which use HTTP are not supported in Reverse Proxy scenarios today. Meanwhile, HTTPS-HTTP bridging, sometimes used in Kerberos Constrained Delegation (KCD) scenarios with TMG/UAG are also not possible as AD FS is HTTPS only.
Extranet soft account lockout imposes an option to temporarily lockout “extranet-connected” accounts, via the Web Application Proxy, by not incrementing the AD BadPassword count on the PDC Emulator in AD once the soft lockout threshold is set in AD FS. If the latter is reached, further logon requests are not passed to AD so that AD Password Policy “hard” lockout measures are not immediately triggered. As the name suggests, this is a soft lockout option that is governed by use of an observation/sliding window that determines how often in a given period a user may attempt to logon via the proxy before the soft count is reached. The goal here is to frustrate password guessing attempts via brute force/DoS from the outside by nefarious users.
Update 29/10 – SamD from the AD product team mentioned that the extranet lockout feature was also done with the view that customers with ADDS account lockout policies can prevent DOS attacks on specific user accounts by setting a threshold lower for the ADFS extranet lockout policy. This way the user still has internal access because ADDS has not locked out the user.
Soft Account Lockout can be invoked through the use of PowerShell.
$observationwindow = New-Timespan -Minutes 1
Set-ADFSProperties –ExtranetLockoutThreshold 3 -EnableExtranetLockout $true -ExtranetObservationWindow $observationwindow
Once set, we can see via Get-ADFSProperties, the changes applied:
ExtranetLockoutThreshold : 3
ExtranetLockoutEnabled : True
ExtranetObservationWindow : 00:02:00
Here we’ve set the lockout threshold to three attempts with an observation window of two minutes.
From a testing standpoint, In the demo AD Domain setup, “hard” account lockout is not set via GPO.
Attempting to login at the RP WIF test application, we’re redirected to AD FS for logon. I enter an incorrect password.
The PDC Emulator FSMO role in AD, which monitors the bad password count (badPwdCount) increments by 1, likewise on the second and third bad password attempts.
I entered a bad password five times in successive attempts. Continued attempts to logon with a bad password, once the observation window kicks in fails to increment the count beyond three for the windowed period of two minutes. Once the window has elapsed, the bad password account is again incremented.
One of the nicer aspects of this setting is that it applies to all endpoints, be they passive or active in nature. This is particularly relevant as it also applies to web services (WS-Trust) endpoints for rich clients, e.g. Office 365.
As covered earlier, devices registered via Workplace Join are registered within Active Directory in the container CN=<Device ID>,CN=RegisteredDevices,DC=mydomain,DC=com. Lost devices can be denied access by disabling or deleting the appropriate object within AD (I moved the device objects to another OU to test this). Access through AD FS is immediately revoked for the workplace joined client.
From testing thus far, devices joined, left and re-registered via Workplace Join are not currently cleaned up within the RegisteredDevices container. Some PowerShell scripting is currently required to enforce this and I would imagine some changes by GA or some scripts made available to manage this process.
Summary
A very long post comes to an end. Next up, we’ll look at UI changes and authentication/access in greater detail and there’s LOTS to cover. As ever, please feel free to comment, contribute, correct and I’ll get back to you!
With the R2 preview of AD FS in Windows Server 2012 out and the large number of changes that are taking place in the new release, I’m going to be bring this post to a quick end; more an abridged version than was originally intended.
In Part I we looked at weaker authentication schemes for step-up scenarios. In Part II, we move onto two-factor/multi-factor authentication (2FA/MFA) use cases.
There’s no real guided approach for doing this in AD FS 2.0, with solutions invariably becoming customized ones, according to the desired use case. Whether AD FS is the authentication provider or occupying a hybrid/broker role, the use of authentication contexts, types and URIs provided by the supported SAML and WS-Federation protocols, become triggers for step-up. Where a context is stipulated, in protocol terms, each is interpreted differently.
SAML supported authentication methods
| Authentication Method | Authentication Context Class URI |
| Username/Password | urn:oasis:names:tc:SAML:2.0:ac:classes:Password |
| Password Protected Transport | urn:oasis:names:tc:SAML:2.0:ac:classes:PasswordProtectedTransport |
| Transport Layer Security (TLS) Client | urn:oasis:names:tc:SAML:2.0:ac:classes:TLSClient |
| X.509 Certificate | urn:oasis:names:tc:SAML:2.0:ac:classes:X509 |
| Integrated Windows Authentication | urn:federation:authentication:windows |
| Kerberos | urn:oasis:names:tc:SAML:2.0:classes:Kerberos |
As we saw in the previous post, forcing authentication in SAML logon scenarios provides support for step-up, but at the expense of single sign-on. In passive federation scenarios, if we don’t specify an authentication method in the request, AD FS will apply authentication according to its own supported methods, attempting to match those against the local authentication types section of the web.config on the AD FS node.
AD FS handles the SAML authentication in order of strength, lowest to highest, from top to bottom as seen in the table: in the default configuration Kerberos is seen as the strongest method. The precedence of the authentication method can be adjusted using the Set-ADFSProperties –AuthenticationContextOrder command to ensure the order meets the requirement we need. The SAML comparison attribute can be used to also influence the authentication method chosen. with AD FS defaulting to a comparison=exact attribute.. As stated in the MSDN article Authentication Handler Overview, if the comparison attribute is set to “better”, “minimum”, or “maximum”, the method of authentication must be stronger than, at least as strong as, or no stronger than one of the specified authentication classes. This gives us some latitude for using it to invoke the desired authentication method through customization of the sign-in process. With SAML the authentication reference is encoded in the SAML request, and to access and to use the information therein, in order to relay the authentication request to our third-party IdP, we need to customize our sign-in process so that the incoming SAML (AuthnContextClassRef) request is decoded. The information we then need needs to be extrapolated so that we can then make the correct routing decision, sending the client to the correct provider. Examples provided on Codeplex hint at how this can be done, by assigning a authentication methods to a relying party using FBA and IWA.
On the other side, let’s look at an example using a WS-Federation based setup.
WS-Federation Passive Profile supported authentication methods
| Authentication Method | Authentication Context (wauth) URI |
| Username/Password | urn:oasis:tc:SAML:1.0:am:password |
| Transport Layer Security (TLS) Client | urn:ietf:rfc:2246 |
| Integrated Windows Authentication | urn:federation:authentication:windows |
With the WS-Federation passive requester profile, the authentication type (wauth) parameter is specified in the query string of the browser or can be specified from the relying party application itself. The whr parameter is used to indicate the claims provide to use for logon.
MFA Step-Up Scenario
Let’s look at a step-up scenario using WS-Federation with an MFA provider.
In the above graphic, we have a third-party MFA provider handling the authentication requests for internal access.. I’ve not included the AD FS Proxy for external access, but should we wish to do so, we could adopt our AD FS proxy configuration and:
– remove the local authentication handlers from the <localAuthenticationTypes> section of web.config to ensure that MFA is always used in external access scenarios;
– follow the same methods used for internal authentication described below;
Internal access step-up scenarios imply the use of the default Integrated Windows Authentication (IWA) handler and the step-up mechanism. With a third-party multi-factor authentication provider, the (MFA) solution is configured within AD FS 2.0 as a claims provider. I’ll use the example of the WIF-based STS from PointSharp which I mentioned in previous posts.
In a step-up context, the STS needs to be able to process the required authentication request and pass this back to AD FS and the RP. With WS-Federation we can do this via the browser as a query string, or, from the web application make use of the wauth and whr parameters to set the authentication method.
Step-Up at the Relying Party Application
Using the WIF 3.5 SDK samples for our relying party, we modify the web.config to specify the desired authentication type. I’ve chosen the X509 certificate handler. For our WIF application this corresponds to an authentication method value of urn:ietf:rfc:2246
<federatedAuthentication>
<wsFederation passiveRedirectEnabled="true" issuer=https://sts.mydomain.com/adfs/ls/ realm=https://rp.mydomain.com/ requireHttps="true" authenticationType="urn:ietf:rfc:2246"
homeRealm=https://2sts.mydomain.com/PointsharpSTS <cookieHandler requireSsl="true" />
</federatedAuthentication>
The homeRealm value is used to specify the claims provider we want to call for step-up, otherwise AD FS itself will attempt to handle the logon, treating as an authentication request for an x509 certificate.
On AD FS we configure the IP-STS as a claims provider.
On the IP-STS, we configure AD FS as a relying party
For convenience, both the STS and AD FS are sharing the same identity store (AD) and I’ve created a test user called demo in Active Directory and a security group called OTP, whom our user is a member of. We’ll use membership of that group on the PointSharp STS as a means of emitting an authentication method claim that corresponds to the expected authentication type. The claims value, urn:ietf:rfc:2246, is returned when a user logs on using MFA (e.g. OTP+password).
The user demo points their browser to the RP URL and with the web.config modifications, the access request is redirected to AD FS and then to the PointSharp STS as the Windows home realm (whr).
On the PointSharp STS during logon, the back-end PointSharp ID services handles the MFA logon request. At the initial prompt, the user enters their user ID and AD password:
They’re challenged to enter an OTP response (hardware/software token or SMS)
Once authenticated, the STS then processes the matching claims rules.
Above, demo is a member of the group OTP, as seen in the claims rule pattern, so a claims type of http://schemas/microsoft.com/ws/2008/06/identity/claims/authenticationmethod is generated. This is used to match the expected reply of the RP with a value urn:ietf:rfc:2246. and this claims is then passed back to AD FS.
In our test web app we see the appropriate authentication method and values being returned. No claims rules on the claims provider or relying parties (at this point) have been done.
Our MFA provider may also support multiple authentication types (AD password, SMS, tokens etc), so the fact that we logon at the STS is not necessarily indicative of a sufficient logon.With AD FS as the RP-STS, we can also block requests that don’t emit the correct authentication method at the STS on the relying party claims pipeline.
c:[Type == "http://schemas.microsoft.com/ws/2008/06/identity/claims/authenticationmethod", Value =~ "^(?i)http://schemas\.microsoft\.com/ws/2008/06/identity/authenticationmethod/tlsclient$"]
=> issue(Type = "http://schemas.microsoft.com/authorization/claims/permit", Value = "PermitUsersWithClaim");
With the authorization rule in place, another user, demo2, that is not a member of the OTP group, when accessing the RP, logs on at the PointSharp STS, but gets the following error at AD FS:
This translates in the event log to an access denied message:
The Federation Service could not authorize token issuance for caller ”. The caller is not authorized to request a token for the relying party ‘https://rp……..’. Please see event 501 with the same instance id for caller identity.
While we can block in this fashion, it makes more sense to enforce and evaluate a more fine-grained access at the application itself. If the app is intended as an internal-only application though, we can also block access from the AD FS proxy via an access rule.
exists([Type == "http://schemas.microsoft.com/2012/01/requestcontext/claims/x-ms-proxy”]) => issue(Type = “http://schemas.microsoft.com/authorization/claims/deny”, Value = “true”);
Step-Up within the Application
The previous steps described step-up at the web application (relying party), rather than within the application itself. The WIF 3.5/4.0 SDK provides samples for step-up authentication using an example of a low-value and a high-value resource within a given application. Here we’re leveraging step-up against a particular resource within an application.
Low-value resources are protected using normal Windows logon credentials (IWA), whereas high-level resources trigger step-up through the use of the certificate handler.
This allows us to add in step-up authentication within the application itself, according to the value of the resource.
A note on Virtual Smart Cards
I mentioned in the last post about virtual smart cards (or VSCs). As an authentication provider, they are available during Windows logon as a credential provider. Given that we wish to expressly enforce step-up at the application, then some sort of adjustment is equired. It may be possible to suppress the use of the VSC as a credential provider and make it available post-logon for application-only access, something that I’ve not tried, but one could argue that this is a case of the tail wagging the dog. We could also use an alternate account as described in Part I or disable single sign-on in the web.config on the AD FS farm to enforce logon between relying parties. The fact that we’re using IWA for normal logon masks somewhat the behaviour of revisiting the AD FS server when switching between IWA-based relying parties.
As ever, please extensively test before considering doing this sort of thing in a production environment.
Alternatives
What’s been discussed so far relates to solutions that are on the “direct” authentication path from an identity federation standpoint. There are also MFA solutions that involve indirect authentication using multi-factor authentication. When I refer to Indirect MFA, Í mean some other component outside of the federation logon process that handles the strong authentication aspects of logon. Diagrams always help to illustrate the cause 
…
1. Via a non-federation capable proxy or gateway component. This component does not integrate with AD FS, but can provide an MFA capability. All connections to the RP are through the proxy. Post-logon, the proxy does some form of credential delegation to the back-end AD FS service. Here’s an example of a logon scenario with Forefront UAG 2010 on a non-federated trunk. While UAG does support federated trunks (as a relying party), MFA on a federated trunk is not (to my knowledge) possible unless we use an upstream claims provider.
2. Via an MFA web agent running on the ADFS proxy as an ISAPI filter or module. This follows the traditional web access management approach of logon via a web agent, with the filter intercepting the call. The user must logon using the stronger form of authentication before the web application, in this case the AD FS proxy pages are exposed. While functionally “integrated”, the two operate technologically independent of one another.
None of the above mechanisms are particularly well suited to step-up because the MFA logon process wraps around normal AD FS logon process, rather than integrating with it. While they may be valid in other access scenarios, i.e. the logon request is intercepted by a handler that doesn’t understand federated logon.
Option 3 is interesting as, via a slight modification, it supports illustrating where MFA is going in the next release of AD FS in Windows Server 2012 R2.
Windows Server 2012 R2
In AD FS R2, Microsoft are making available an MFA API for vendors to plug-in to and integrate with the federation service directly, allowing for a more rich logon experience. This also extends the concept of using multi-factor authentication not only at a protocol and application level, but at a policy level too, whereby policy per relying party for authentication is achievable. It also suggests that the redirect method via claims providers is being eschewed by Microsoft in favour of a more AD FS centric approach. I suspect the aim here is to provide a more level playing field for integrating external authentication providers with AD FS rather than placing the burden on the vendor and heavy use of customization to support scenarios described..
What this means for step-up scenarios we shall see. I’ll be looking at this raft of new features with AD FS R2 in future posts.
The Access Onion 