by Mitja Kolsek, the 0patch Team
Forced authentication issues (including NTLM relaying and Kerberos relaying) are a silent elephant in the room in Windows networks, where an attacker inside the network can force a chosen computer in the same network to perform authentication over the network such that the attacker can intercept its request. In the process, the attacker obtains some user's or computer account's credentials and can then use these to perform actions with the "borrowed" identity.
In case of PetitPotam, for instance, the attacker forces a Windows server to authenticate to a computer of their choice using the computer account - which can lead to arbitrary code execution on the server. With RemotePotato0, an attacker already logged in to a Windows computer (e.g., a Terminal Server) can force the computer to reveal credentials of any other user also logged in to the same computer.
For a great primer on relaying attacks in Windows, check out the article "I’m bringing relaying back: A comprehensive guide on relaying anno 2022" by Jean-François Maes of TrustedSec. Dirk-jan Mollema of Outsider Security also wrote several excellent pieces: "The worst of both worlds: Combining NTLM Relaying and Kerberos delegation", "Exploiting CVE-2019-1040 - Combining relay vulnerabilities for RCE and Domain Admin" and "NTLM relaying to AD CS - On certificates, printers and a little hippo."
Alas, Microsoft's position seems to be not to fix forced authentication issues unless an attack can be mounted anonymously; their fix for PetitPotam confirms that - they only addressed the anonymous attack vector. In other words:
If any domain user in a typical enterprise network should decide to become domain administrator, no official patch will be made available to prevent them from doing so.
Microsoft does suggest (here, here) various countermeasures to mitigate such attacks, including disabling NTLM, enabling EPA for Certificate Authority, or requesting LDAP signing and channel binding. These mitigations, however, are often a no-go for large organizations as they would break existing processes. It therefore isn't surprising that many of our large customers ask us for micropatches to address these issues in their networks.
Consequently, at 0patch we've decided to address all known forced authentication issues in Windows exploitable by either anonymous or low-privileged attackers.
The Vulnerability
The vulnerability we micropatched this time has two names - PrinterBug and SpoolSample - but no CVE ID as it is considered a "won't fix" by the vendor. Its first public reference is this 2018 Derbycon presentation "The Unintended Risks of Trusting Active Directory" by
Will Schroeder, Lee Christensen, and Matt Nelson of SpecterOps, where authors describe how the MS-RPRN RPC interface can be used to force a remote computer to initiate authentication to attacker's computer.
Will Schroeder's subsequent paper "Not A Security Boundary: Breaking Forest Trusts" explains how this bug can be used for breaking the forest trust relationships; with March 2019 Windows Updates, Microsoft provided a related fix for CVE-2019-0683, addressing only the forest trust issue.
Today, four-plus years later, the PrinterBug/SpoolSample still works on all Windows systems for forcing a Windows computer running Print Spooler service to authenticate to attacker's computer, provided the attacker knows any domain user's credentials. As such, it is comparable to PetitPotam, which also still works for a low-privileged attacker (Microsoft only fixed the anonymous attack), and the recently disclosed DFSCoerce issue - which we're also preparing a micropatch for.
The vulnerability can be triggered by making a remote procedure call to a Windows computer (e.g., domain controller) running Print Spooler Service, specifically calling function RpcRemoteFindFirstPrinterChangeNotification(Ex) and providing the address of attacker's computer in the pszLocalMachine argument. Upon receiving such request, Print Spooler Service establishes an RPC channel back to attacker's computer - authenticating as the local computer account! This is enough for the attacker to relay received credentials to a certificate service in the network and obtain a privileged certificate.
When RpcRemoteFindFirstPrinterChangeNotification(Ex) is called, it impersonates the client via the YImpersonateClient function - which is good. The execution then continues towards the vulnerability by calling RemoteFindFirstPrinterChangeNotification. This function then calls SetupReplyNotification, which in turn calls OpenReplyRemote: this function reverts the impersonation (!) before calling RpcReplyOpenPrinter, where an RPC request to the attacker-specified host is made using the computer account.
We're not sure why developers decided to revert impersonation of the caller before making that RPC call, but suspect it was to ensure the call would have sufficient permissions to succeed regardless of the caller's identity. In any case, this allow the attacker to effectively exchange low-privileged credentials for high-privileged ones.
Our Micropatch
When patching an NTLM relaying issue, we have a number of options, for instance:
- using client impersonation, so the attacker only receives their own credentials instead of server's,
- adding an access check to see if the calling user has sufficient permissions for the call at all, or
- outright cutting off the vulnerable functionality, when it seems hard to fix or unlikely to be used.
This particular bug fell into the latter category, as we could not find a single product actually using the affected functionality, and Windows are also not using it in their printer-related products. If it turns out our assessment was incorrect, we can easily revoke this patch and replace it with one that performs impersonation.
Our micropatch is very simple: it simulates an "access denied" (error code 5) response from the RpcReplyOpenPrinter function without letting it make the "leaking" RPC call. This also blocks the same attack that might be launched via other functions that call RpcReplyOpenPrinter.
Source code of the micropatch has just two CPU instructions:
MODULE_PATH "..\Affected_Modules\spoolsv.exe_10.0.17763.2803_Srv2019_64-bit_u202205\spoolsv.exe"
PATCH_ID 908
PATCH_FORMAT_VER 2
VULN_ID 7419
PLATFORM win64
patchlet_start
PATCHLET_ID 1
PATCHLET_TYPE 2
PATCHLET_OFFSET 0x576cc
N_ORIGINALBYTES 5
JUMPOVERBYTES 0
PIT spoolsv.exe!0x577df
; 0x577df -> return block
code_start
mov ebx, 5
jmp PIT_0x577df
code_end
patchlet_end
Micropatch Availability
While this vulnerability has no official patch and could be considered a "0day", Microsoft seems determined not to fix relaying issues such as this one; therefore, this micropatch is not provided in the FREE plan but requires a PRO or Enterprise license.
The micropatch was written for the following Versions of Windows with all available Windows Updates installed:
- Windows 11 v21H2
- Windows 10 v21H2
- Windows 10 v21H1
- Windows 10 v20H2
- Windows 10 v2004
- Windows 10 v1909
- Windows 10 v1903
- Windows 10 v1809
- Windows 10 v1803
- Windows 7 (no ESU, ESU year 1, ESU year 2)
- Windows Server 2008 R2 (no ESU, ESU year 1, ESU year 2)
- Windows Server 2012
- Windows Server 2012 R2
- Windows Server 2016
- Windows Server 2019
- Windows Server 2022
This micropatch has already been distributed to, and is being applied to, all online 0patch Agents in PRO or Enterprise accounts (unless Enterprise group settings prevent that).
If you're new to 0patch, create a free account in 0patch Central, then install and register 0patch Agent from 0patch.com, and email sales@0patch.com for a trial. Everything else will happen automatically. No computer reboot will be needed.
We'd like to thank Will Schroeder, Lee Christensen, and Matt Nelson of SpecterOps for sharing details about this vulnerability, and Dirk-jan Mollema of Outsider Security for excellent articles on relaying attacks and exploiting PrinterBug/SpoolSample in particular. We also encourage security researchers to privately share their analyses with us for micropatching.
No comments:
Post a Comment