HTTP Strict Transport Security (HSTS) is a feature commonly used by websites to protect against protocol downgrade attacks, where an attacker forces the use of insecure HTTP instead of HTTPS. If successful, the attacker can intercept unencrypted communication between the client and the web server. This is undesirable for obvious reasons. As such, web server administrators implement an HTTP response header named Strict-Transport-Security with some additional settings that instruct the user agent, in this case, a web browser, to only use secure HTTPS when communicating with the web server. Attempts to use HTTP will not work.
VPN and SSTP
As security is always a top concern when building an Always On VPN infrastructure, careful attention must be paid to VPN protocol configuration to ensure optimal security. Secure Socket Tunneling Protocol (SSTP) is a popular VPN protocol for Always On VPN user tunnel connections. SSTP uses Transport Layer Security (TLS) for encryption, so administrators are encouraged to implement recommended security configurations, such as disabling insecure protocols like TLS 1.0 and TLS 1.1 and optimizing TLS cipher suites as described here.
SSTP with HSTS
It would seem that enabling HSTS on a Windows RRAS VPN server would be ideal for improving SSTP security. However, that’s not the case. HSTS prevents protocol downgrade attacks from HTTPS to HTTP, but SSTP already uses HTTPS exclusively, making the use of HSTS irrelevant. If an attacker attempted a protocol downgrade attack on an SSTP VPN connection, it would fail because the service does not support HTTP between the client and the VPN gateway. Additionally, even if it were possible to configure RRAS to send an HSTS response header, it would be ignored by the client because the user agent is not a web browser.
Microsoft introduced important changes affecting certificate-based authentication on Windows domain controllers as part of the May 10, 2022 update KB5014754 that may affect Always On VPN deployments. The update addresses privilege escalation vulnerabilities when a domain controller is processing a certificate-based authentication request. The recommendation from Microsoft is that the update be applied to all Windows domain controllers and Active Directory Certificate Services (AD CS) servers as soon as possible.
After applying the update to certification authority (CA) servers, a non-critical extension with Object Identifier (OID) 1.3.6.1.4.1.311.25.2 is added to all issued certificates with the user or device security identifier (SID) included. Domain controllers with the update installed will use this information to validate the certificate used for authentication and ensure that it matches the information in Active Directory.
Domain Controllers
The update operates in Compatibility Mode, by default, when applied to domain controllers. Windows monitors authentication requests and records audit events for certificates presented for authentication under the following conditions.
No strong mapping (event ID 39) – The certificate has not been mapped explicitly to a domain account, and the certificate did not include the new SID extension.
Certificate predates account (event ID 40) – A certificate was issued before the user existed in Active Directory, and no explicit mapping could be found.
User’s SID does not match certificate (event ID 41) – A certificate contains the new SID extension, but it does not match the SID of the corresponding user account.
Certificate Mapping
Administrators can map certificates explicitly to accounts in Active Directory, but this results in a significant administrative burden in most environments. A better option is to reissue user and device authentication certificates after applying the KB5014754 update to all issuing CA servers.
Reenroll Certificates
Administrators should reissue user and device authentication certificates after applying the KB5014754 update. Open the Certificate Templates management console (certtmpl.msc), identify the user or device authentication certificate template, then right-click on the template and choose Reenroll All Certificate Holders.
Enforcement Mode
After applying update KB5014754, administrators should monitor domain controller event logs for event IDs 39, 40, and 41. Once all certificates have been updated, and none of these events have been recorded for 30 days, administrators can switch to Full Enforcement Mode by enabling it in the registry on all domain controllers.
Updated 12/8/2022: Microsoft has pushed back the original enforcement date of May 9, 2023, to November 14, 2023 “or later”. Stay tuned!
Updated 11/29/2022: Microsoft now states that the full enforcmenet mode date is now February 11, 2025 “or later”. Once again, stay tuned!
Known Issues
There have been some reports of authentication issues after installing the KB5014754 update. Early indications are that device authentication certificates missing a Subject Alternative Name (SAN) entry are to blame. Administrators are encouraged to update their device certificates to include the SAN entry. Optionally, but not recommended, administrators can place the update in disabled mode by editing the registry.
Note: An out-of-band update for these authentication issues is now available. See the reference links at the top of this article for more information.
Caveat
It’s important to understand that this new OID is added only to online templates. Online templates are those that build the subject information from Active Directory. Unfortunately, this new OID is NOT applied to offline templates (templates where the subject name is supplied in the request), such as those used for delivering certificates with Microsoft Endpoint Manager/Intune using PKCS or SCEP. It is impossible to move to enforcement mode when issuing user or device authentication certificates with Microsoft Endpoint Manager or Intune today. Microsoft is aware of this limitation and is working to address this issue as we speak. I expect a fix to be available sometime before the May 2023 deadline when Microsoft permanently switches on enforcement mode.
Always On VPN Administrators may encounter a scenario where Always On VPN connections suddenly stop working for all clients using the Secure Socket Tunneling Protocol (SSTP) VPN protocol. IKEv2 VPN connections continue to work, however.
Event Log
Reviewing the event log on a client machine reveals an error event ID 20227 from the RasClient source. The error message states the following.
“The user [username] dialed a connection named [connection name] which has failed. The error code returned on failure is -2146762495.”
Error -2146762495?
Always On VPN administrators will be familiar with error codes such as 809, 691 and 812, 853, 858, and even 13801, 13806, and 13868. However, this error code seems to be formatted much differently. As it turns out, this message is in decimal format. Thankfully it’s pretty easy to convert it to something more meaningful, like hexadecimal. To do this, open the Windows calculator (calc.exe) and switch to programmer mode. Highlight DEC and enter -2146762495. The hexadecimal value will be displayed in the HEX field, as shown here.
Error 0x800B0101
After converting the error message from decimal to hex, use the Microsoft Error Lookup tool (err.exe) to translate the hex value of this error. As shown here, 0x800B0101 translates to CERT_E_EXPIRED.
Expired TLS Certificate
Once again, an expired certificate is to blame! In this case, the TLS certificate installed on the VPN server has expired and is no longer valid.
Resolution
The problem is simple enough to resolve, of course. Obtain a new TLS certificate from your certification authority (CA) of choice and update your VPN server configuration. You can find detailed guidance for updating the RRAS VPN server’s TLS certificate here. You will also find a video demonstration of the RRAS SSL/TLS certificate renewal process here.
