Windows RRAS VPN and Multi Factor Authentication

Posted on 6 CommentsPosted in Azure, MFA, multi-factor auth, password, phone factor, policy, pptp, remote desktop, rras, sdk, vpn

This blog post covers the steps to add Multi Factor Authentication (MFA) to Windows RRAS server. Once this is enabled, and you sign in with a user enabled for MFA in Azure Multi-Factor Authentication Server (an on-premises server) you are required to answer your phone before you can connect over the VPN. That is, you connect to the VPN endpoint, enter your username and password and if they are correct, then confirm that you want to authenticate by answering your phone. If you are not connecting over VPN and someone else is and using your credentials, unless they also have your phone they are not going to succeed! And all this for less than a £1 per user per month!

This configuration requires the following components set up:

  • Multi Factor Authentication set up in Azure
  • Azure Multi-Factor Authentication Server installed on-premises
  • Some users configured in Azure Multi-Factor Authentication Server
  • RRAS VPN server configured to use RADIUS for authentication, with the MFA server being the RADIUS endpoint

Step 1: MFA setup in Microsoft Azure

To do this you need an Azure subscription and DirSync configured to populate the Azure Active Directory with users. If you already have Office 365 with DirSync then you have this configuration already and you can login to Azure using the Azure AD link from the Office 365 management portal.

Once in Azure select “Active Directory” from the portal and click “Multi-Factor Auth Providers” from the menu at the top. You will probably not have any providers listed here, but if you do already (for example you are already using MFA for Office 365 or AD FS) then you can use the existing provider. To add a provider click Add, select “Multi-Factor Auth Provider” and “Quick Create” as shown:

image

Provide a name and then choose a usage model. Usage models are per user or per authentication. Per User works when a single user will authenticate more than 10 times a month. When users would only use MFA occasionally you can buy the service by the authentication request. For example if you had 200 VPN users who connected each day, you would choose Per User. But if you had 200 VPN users, who only dialled in once a month (i.e. a total of 200 authentications) then you would be better off buying the Per Authentication model as you would pay for 20 batches of authentications (each batch allows 10 authentications regardless of the user). You cannot change the authentication model without removing the auth provider and making a new one.

Finally, link the provider to your directory.

Select your auth provider once it is created and click Manage at the bottom of the portal:

image

This opens a new tab in the browser and takes you to the Azure Multi-Factor Authentication management pages.

Whilst here, as there is actually not a lot to do here, take a look at Configure to see what settings you can change. Maybe enter your email address for the fraud alert notifications, but leave everything else as is for now.

Back on the home page of the Azure Multi-Factor Authentication web site, click Downloads.

Step 2: Installing Multi-Factor Authentication Server

From the Downloads page find the small download link (above the Generate Activation Credentials button) and download the software to a Windows Server that is joined to your domain.

On the said server install .NET 2.0 and IIS with the default settings. Ensure that you have a digital certificate installed, as the web site the the users will go to for provisioning and managing their device is available over SSL. Mobile phones can use the app to validate connections as well, and that will be the subject of a different blog post, but you need a trusted cert that is valid and has a subject name such as mfa.domain.com (where domain.com is your domain) and so a 3rd party cert is required. In this blog I have used my wildcard cert from DigiCert.

Run the Multi-Factor Authentication Server installer and proceed through the steps. Use the wizard to configure the server and select VPN. During the installation you will also need to authenticate the Multi-Factor Authentication Server to Azure. This requires a set of credentials that are valid for ten minutes at a time, and generated from the Generate Activation Credentials button in the management web page at Azure. So don’t click this button until the Multi-Factor Authentication Server requires this info.

For this blog I am going to protect my VPN with Azure MFA. Therefore during the configuration wizard I select just the VPN option:

image

As you proceed through the wizard you will be asked about the RADIUS client configuration needed for your VPN provider. In here enter the IP address of your RRAS box and a password that you have made up for the occasion. You will need this password, or shared secret, when configuring the RRAS server later.

image

Finish the installation of Multi-Factor Authentication Server.

Once complete, open the Multi-Factor Authentication Server management program and select RADIUS Authentication. Ensure Enable RADIUS authentication is selected as this will allow this server to provide authentication on behalf of the RADIUS client and therefore insert requests for MFA via the users phone into the authentication flow.

image

Double click the IP address of your VPN server and select “Require User Match”

Step 3: Configure Users for MFA

Click the Users icon in Multi-Factor Authentication Server and click Import from Active Directory. Set the filtering to add just the users you want to enable MFA for. A user who dials in who is not listed here will not be blocked from authentication to the VPN.

image

A user will have a yellow warning icon next to it if it is disabled. For disabled users you can either allow authentication to pass through the MFA server without requiring the user to have the second factor of authentication working. This can be set on the users properties, and the Advanced tab by selecting Succeed Authentication for “When user is disabled”. The enabled check box is on the general tab.

If a user is enabled here then they will need to either complete the MFA authentication process. The exact process the user needs to do to pass the authentication process always starts with getting their username and password correct. After that they can do one of the following:

  • Press # when the call comes through to their phone
  • Reply to a text message – texts go to a US number, so this might cost the user international rates!
  • Press the Verify button on the MFA app on their phone
  • Optionally add a PIN number to any of the above – for example, when the MFA call comes through to enter your PIN and then press # rather than just #.

Each user can have different settings. When you import users from the Active Directory it reads (by default) their mobile number from the Active Directory as the primary number to authenticate against. You can set backup numbers if required. If a user has a mobile number they are enabled by default. When importing you can set which MFA method the user will use, and you can install the MFA portal so the user can change their own settings if you want (outside the scope of this blog).

By now you have Azure MFA configured, the MFA server installed on-premises (it will need port 443 access to Azure to complete the authentication) and users set up in the MFA server. The MFA server is also configured to act as a RADIUS endpoint for your VPN service. If you install more than one MFA server for load balancing and HA, ensure that each MFA server is selected on the Multi-Factor Auth Servers tab on the RADIUS settings – this starts the MFA RADIUS service on each selected machine.

Before you configure VPN, final step here is to test the user. From the Users area on the MFA server select a user and click Test. Authenticate as the user, username and password required for this test, and then press # after answering the phone. Try out the SMS or text message form factor for authentication as well. To support the mobile app you need to install the users portal, the SDK and the mobile app web service – so thats for a different blog post.

Step 4: Configure RRAS VPN to Use Multi-Factor Authentication

Finally, change to your RRAS server. Before going any further, ensure that RRAS is working before MFA is enabled – you don’t want to troubleshoot MFA only to find it was RRAS not working in the first place! The RRAS server’s IP address must match the IP address listed under the RADIUS configuration in the MFA server.

Right-click the RRAS server name in the Routing and Remote Access console. If you are setting up MFA for another type of VPN server then any that supports RADIUS will do. In the server properties, select the Security tab and change the Authentication provider to RADIUS Authentication (it was probably Windows Authentication).

image

Click Configure to the right of this drop-down and click Add:

image

Enter the IP address of your MFA server, repeating the Add process if you have more than one MFA server configured. Enter the shared secret that you used when setting up the MFA server and ensure that the timeout is set to 60 seconds. This is an important setting. When the user connects to the VPN server, the timeout needs to exceed the time it will take for the users phone to ring, listen the the greeting, enter the PIN (optionally) and press #. One minute should be enough to do this. After one minute the RRAS VPN server will automatically fail authentication, so the user has one minute to complete the second factor authentication on their phone.

You should now be able to dial into your VPN and authenticate with your username and password. Once you succeed with this, the MFA authentication starts and the call will arrive on your phone:

image

You can get the graphic as a vCard from http://1drv.ms/1xXCA01. Download this vCard, save it to your contacts and when you sync your contacts to your phone, your phone will tell you the Microsoft Phone Auth service is calling. You could change the name and graphic to suit, just make sure the number matches the CallerID setting in Azure MFA.

Whilst you are waiting for the call the arrive, and before you accept the auth request, the VPN client appears to pause:

image

Once you complete the auth, the VPN session starts up. If the call and time to answer exceeds 60 seconds, then consider increasing the RADIUS timeout on the VPN server.

Finally, and this will be a different blog post, you might want to offer the user a portal they can go to to change their settings such as updating phone number and changing mode of authentication etc. But this is off topic for this post. Later posts will cover using this MFA server integrated with AD FS and OWA as well.

Exchange 2013 Transport Agents

Posted on 3 CommentsPosted in 2007, 2010, 2013, agent, exchange, sdk, smtp, visual studio

Earlier today I posted http://blog.c7solutions.com/2012/10/creating-simple-exchange-server.html on how to create a transport agent in Exchange, and though the steps cover some of the detail for Exchange Server 2013 they do not cover some of the detail, so I’ve added that to this post below:

  • Use .NET Framework 4.0 for Exchange 2013 though it is possible to write the agents using earlier versions of the .NET Framework. If you do use earlier versions then you need to update some of the config files on the server to state the version levels that you support and Enable Support for Legacy Transport Agents on Technet lists the steps for this.
  • The Front End Transport role (which is installed on the CAS Server role) can support SMTP agents bound to events up to, but not including OnEndOfData.
  • The Hub Transport role (which is installed on the Mailbox Server role) can support all SMTP agents except for those that bind to the OnConnect event.
  • The Hub Transport role also support routing agents and delivery agents. The CAS role supports neither of these agent types.
  • The Exchange 2013 Edge Transport Server role (expected with Exchange 2013 SP1) will support agents at the OnConnect event.
  • To install a transport agent on a multi-role server (CAS and Mailbox on the same machine) then you need to use can use -TransportService Hub to the Install-TransportAgent cmdlet shown in the earlier blog if you want to bind the agent to the Hub Transport events. If you want to use the Front End Transport events (OnMailFrom to OnEndOfHeaders) then use –TransportService FrontEnd instead (updated 2 Oct from Philippe comment below).
  • To install a transport agent on a CAS only server role you need to use local PowerShell and load the Exchange Management Shell snap-in with Add-PSSnapin Microsoft.Exchange.Management.PowerShell.E2010
  • Finally, to view the agents in a given service, always state the service name or you will get back the Hub Transport service by default. For example Get-TransportAgent –TransportService FrontEnd for the agents bound to the Front End Transport service. Other values are Hub, Edge and MailboxSubmission and MailboxDelivery (though the last two don’t seem to work in the pre-release version on Exchange Server 2013.

Creating a Simple Exchange Server Transport Agent

Posted on 5 CommentsPosted in 2007, 2010, 2013, agent, exchange, mcm, sdk, smtp, transport agent, visual studio

This blog post follows a session that I delivered at the MEC 2012 conference in Orlando. If you attended the conference the slides are available on http://mymec.mecisback.com for the rest of 2012.

Part of the transport agents session was writing a new transport agent, and the example agent was to do add a form of catch-all functionality. The example code below (which purposefully needs editing to work, as its designed to be a learning tool) takes an email in the form of alias_tag@domain.com where the final recipient is alias@domain.com and the subject line is changed to read [tag] Original Subject. Therefore email addresses can be given out that are an adjusted form of your real address, and on receipt of the email the correct recipient address is determined in code (the value before the _ appended to the domain name) so that you do not need to add lots of aliases to your user account in Active Directory. The subject is then changed to indicate the email came from this alternative original address. For example, an email to bill_linkedin@microsoft.com would go to bill@microsoft.com and have [linkedin] added to the start of the subject.

As this partial catch all transport agent needs access to the recipient information and the subject, the agent needs to be bound to OnEndOfHeaders or OnEndOfData in the SMTP stack. It could also be bound to OnSubmittedMessage as a routing agent (rather than an SMTP agent) but could not be bound to OnResolvedMessage as recipient resolution has already happened by this point.

Writing the Transport Agent

To create the transport agent you need Visual Studio – the basic version of Visual Studio are sufficient and you need to copy the two DLL’s from Program Files\Microsoft\Exchange Server\V14\Public (or the V15 folder for Exchange 2013) to a folder on your development machine. You need a copy of these DLL’s for every version/service pack/update rollup of Exchange that you will run your agent on, as the DLL’s might change between these versions, and if you have the wrong version you will not be able to install your agent on a new server or if the server is updated, the transport service will fail as it will not be able to load the agent.

In Visual Studio, create a .NET 2 or 3.5 Class Library (for Visual Basic if using the below code) for Exchange 2010 or .NET 4.0 Class Library for Exchange 2013. For the project name enter something descriptive for what you are going to create (rather then ClassLibrary1). For example MECDemoPartialCatchAll.

image

Copy and paste the following code into the class.vb file, replacing the template text.

   1: Imports System

   2: Imports System.Collections.Generic

   3: Imports System.Text

   4: Imports Microsoft.Exchange.Data.Transport

   5: Imports Microsoft.Exchange.Data.Transport.Smtp

   6:

   7: Namespace XXXXX REM Change This

   8:

   9:     NotInheritable Class YYYYYY REM Change This

  10:         Inherits SmtpReceiveAgentFactory

  11:

  12:         Public Overrides Function CreateAgent(ByVal server As SmtpServer) As SmtpReceiveAgent

  13:             Return New ZZZZZ REM Change This

  14:         End Function

  15:

  16:     End Class

  17:

  18:     Public Class ZZZZZ REM Change This

  19:         Inherits SmtpReceiveAgent

  20:

  21:         Private Sub MyEndOfDataHandler(ByVal source As ReceiveMessageEventSource, ByVal e As EndOfDataEventArgs) Handles Me REM Change This

  22:             ' Get and change the recipient from alias_tag to alias (only doing for 1 recipient for simplicity)

  23:             If e.MailItem.Recipients.Count = 1 And InStr(e.MailItem.Recipients.Item(0).Address, "_") > 1 Then

  24:                 Dim Recipient() As String = Split(e.MailItem.Recipients.Item(0).Address, "@", 2)

  25:                 Dim EmailAlias() As String = Split(Recipient(0), "_", 2)

  26:                 'EmailAlias(0) = alias to send email to

  27:                 'EmailAlias(1) = tag for subject line

  28:                 'Recipient(1) = domain

  29:

  30:                 ' The following line prepends [tag] to the subject of the message.

  31: REM Change This                 "[" + EmailAlias(1).ToString + "] " + e.MailItem.Message.Subject

  32:

  33:                 'the following drops the current recipient

  34:                 e.MailItem.Recipients.Remove(e.MailItem.Recipients.Item(0).Address)

  35:

  36:                 'the following adds the recipient back again, this time using the alias without the tag

  37:                 e.MailItem.Recipients.Add(New RoutingAddress(EmailAlias(0).ToString + "@" + Recipient(1).ToString))

  38:             End If

  39:         End Sub

  40:

  41:     End Class

  42:

  43: End Namespace

Download this code

The remaining steps on creating the agent will be to modify the lines above that are marked with REM statements and then to add the reference DLL’s you copied from your Exchange Server and finally build your DLL.

Visual Studio can tell you when your code contains errors before you build your code, but to do so you need to reference the DLL’s that you obtained from your Exchange Server earlier in this blog. To do this you need to have copied the two DLL’s to a unique folder on your computer. I use c:\temp\Agent Authoring\Dll\E14-SP2-RU4 to store the DLL’s from Exchange 2010, SP2, RU4. Then when RU5 is released, if the DLL’s have changed I will place them in an E14-SP2-RU5 folder and update the references in my project. If I keep using the same folder then Visual Studio does not refresh the DLL but uses an existing cached copy.

To add the DLL’s right-click the project name to the right of the Visual Studio window and choose Properties (or press Alt+Enter):

image

Change to the Reference’s tab and click Add > Browse to find and add these two DLL’s.

image

You should see the two DLL’s listed as well as the default references for DLL’s included in the class library template that you used in Visual Studio.

Back on the code, its time to make the changes. Each of the changes and why it is needed are detailed based on the line numbers above

Line 7 – Namespace

This value of the namespace for the transport agent. Standard conventions indicate that this should be your company name. Therefore in your code change XXXXX to C7Solutions

Line 9 – NotInheritable Class

This value, YYYYYY, is the class that you are creating. This class inherits all the functionality of the Microsoft.Exchange.Data.Transport.Smtp.SmtpReceiveAgentFactory class. For this partial catch-all agent, a good name would be PartialCatchAllFactory.

Line 13 and 18 – Public Class

The Public Class contains the code to execute when the agent is called. This class has a name (ZZZZZ in the sample code above) and should be given a name that represents what the code does. This name in this example can be PartialCatchAll and it needs to be used in Line 13 (inside the PartialCatchAllFactory code to indicate the code to call instead of SmtpReceiveAgentFactory. And of course it is needed on line 18 to name the actual block of code. The name to use for this example will be PartialCatchAll

Line 21 – Handler

This line is missing the end of the code. Its a line that says run the code in this subroutine if the OnEndOfData event is called. Change the end of the line to read Me.OnEndOfData. This value should be one of the suggestions in the drop down list if you have done everything correct so far:

image

This subroutine also returns e as a pointer to the email message (i.e. you can modify e.MailItem.DeliveryPriority and many other properties) and source as a reference to the SMTP connection (i.e. with source.Disconnect you would close the SMTP session).

In this code lines 34 and 37 change the recipient. Line 34 removes the first recipient and then line 37 adds a new recipient that matches the alias + the domain.

Line 31 – Modify Subject

Finally, this code is missing the start of line 31. You need to enter e.MailItem.Message.Subject =  at the start of the line so that they subject becomes [tag] + original subject.

Building the Transport Agent DLL

The Errors List at the bottom of the Visual Studio screen should be empty by now. So time to build the project. If you have not saved this project in Visual Studio, the DLL will be created in the %temp% directory, so best recommendation is to save the project before building the DLL.

To save the project click File >  Save All and enter a suitable name for the project.

image

The project name becomes the solution name by default and then click Save.

The final thing to check before you build the DLL is the Root namespace value. This should be set to the name of the class library and can be set during creation of the library. If you have written your library in VB.Net then you will need this value to register the DLL. If you used C# then you will not need this value. As we have used VB.Net above, we need to check the root namespace value. This can be found on the project properties page.

image

If your DLL is ready to release (i.e. you have build and tested it already) then choose Build > Configuration Manager menu. Change the Active solution configuration to Release and click Close. If you do not make this change then the DLL will be created in the project_name\bin\debug folder and its possible that Visual Studio does not compile all the optimizations to the code that it can. For production items that perform at as good a speed as you can write optimal code, you should change this to Release for the Configuration value. The DLL will be created in project_folder\bin\release.

To build your DLL (debug or release) select the Build > Build RootNamespaceValue menu. This will create a DLL in project_name\bin\debug or project_name\bin\release.  A working version for Exchange 2010 SP2, RU4 can be downloaded from here.

Installing the Transport Agent

Copy the DLL to all your Exchange Servers that have the hub transport role installed on them (or if this should only apply to emails inbound from the internet, then just the servers that are listed on your MX records or the first servers connected to for inbound emails). Place the file in a directory (say C:\C7Agents) and run the following five lines from Exchange Management Shell:

Install-TransportAgent -Name “C7PartialCatchAll” -AssemblyPath “C:\C7Agents\MECDemoPartialCatchAll.dll”
-TransportAgentFactory “MECDemoPartialCatchAll.C7Solutions.PartialCatchAllFactory”

Enable-TransportAgent “C7CatchAll”

Restart-Service MSExchangeTransport

IISRESET

# Recommend closing EMS window now

Note that this restarts the transport service (required) and IIS (which will effect OWA and other CAS roles on the same machine). Remote Powershell (an IIS resource) will lock the DLL open, and so if you need to delete the DLL after uninstalling it, you need to have reset IIS and closed the Powershell window.

Finally, send an email to name_value@yourdomain.com where name is the part of your email address before the @ in Exchange and yourdomain.com is your domain. The value in “value” will be added to the subject as [value]. You get get an email in this form you know your agent has worked.

Transport Agents and Exchange 2013

I’ve also written an additional blog post on the changes for Exchange Server 2013 and transport agents. This covers the things to consider that are different with regards to Exchange 2007/2010 and the new Exchange.