← cd ../blog

CVE-2026-124194: an Outlook message file that leaks your NTLM hash without a warning

An exploited Outlook spoofing bug (CVSS 6.5, CISA KEV) skips the external-content warning and leaks the user's NTLMv2 hash when a crafted file is opened. Fixed builds, egress controls and hunting.

CVE-2026-124194 is one of November’s Patch Tuesday fixes for Outlook. Microsoft marked it “Exploitation Detected,” and CISA added it to the KEV catalog the same day, 10 November 2026. The score is only 6.5 Medium, which is exactly why I’m writing it up: scanners and patch dashboards will rank it below a pile of 8s and 9s, but it’s a confirmed in-the-wild credential theft primitive, and those tend to be the first step of something larger. One thing to get straight early, because “Outlook zero-day” headlines invite the assumption: this is not a zero-click, Preview Pane bug. Microsoft says the user has to open the file.

Scope and ethics. This post covers how the bug works at a high level, exposure checks, detection and remediation for systems you own or are authorized to assess. It contains no crafted messages, file structures or trigger details. Lab validation means checking builds, egress rules and NTLM auditing, not sending lures to anyone.

At a glance

   
CVE CVE-2026-124194, “Microsoft Outlook Spoofing Vulnerability”
Severity CVSS 3.1 6.5 Medium, AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N (Microsoft, via NVD). Microsoft’s temporal score marks exploit maturity as High (E:H)
CWE CWE-356 Product UI does not Warn User of Unsafe Actions
Affected Microsoft 365 Apps for Enterprise, Office LTSC 2021, Office LTSC 2024 (32- and 64-bit)
Fixed in Office LTSC 2021 16.0.14334.21048, Office LTSC 2024 16.0.17932.21140, Microsoft 365 Apps per channel (table below)
Exploited? Yes. Microsoft: “Exploitation Detected” (not publicly disclosed beforehand). CISA KEV, added 10 Nov 2026 (federal due date 24 Nov)
Auth / interaction None / user must open a crafted file. The Preview Pane is not an attack vector

Description: how the bug works

Outlook has long had a safety rule: before it fetches content from an external location, such as a remote image or a linked resource, it warns the user or blocks the fetch. That rule matters on Windows in particular, because when Windows connects to a remote UNC path (\\server\share), it can automatically try to authenticate with the logged-on user’s credentials using NTLM. Whoever runs that server receives an NTLMv2 challenge-response.

Microsoft’s FAQ describes CVE-2026-124194 this way: an attacker sends “a specially crafted message file” and convinces the user to open it in Outlook. Opening it “could cause Outlook to connect to an attacker-controlled network location without displaying the usual external-content warning.” That’s CWE-356 exactly. The dangerous action itself isn’t new. The bug is that the safeguard which should have stopped it, or at least asked first, doesn’t fire for this content.

Microsoft hasn’t said which element of the file triggers the connection, and I won’t speculate. Note a small inconsistency in Microsoft’s own FAQ: one answer says “message file” and another says “malicious Office file.” Either way, it reaches the user as a file they open.

Why “spoofing”? Microsoft’s impact category describes what the attacker does with the result. The FAQ: the bug “discloses a user’s NTLMv2 hash to the attacker who could use this to authenticate as the user.” In practice, a captured NTLMv2 response can be cracked offline if the password is weak, or relayed in real time to another service that accepts NTLM.

Microsoft credits Benoît Sevens (“using Claude with Anthropic”) and a researcher named “nono.” Microsoft hasn’t named the threat actor, targets or campaign scale.

Impact

  • Credential exposure for whoever opens the file: their NTLMv2 response goes to the attacker’s server.
  • Account takeover if the hash is cracked, or NTLM relay to services without signing or Extended Protection for Authentication (EPA). Relay against things like SMB without signing, LDAP without channel binding, or AD CS web enrollment has been a well-worn path to domain compromise for years.
  • Low barrier. The CVSS vector has no privileges and low complexity. The only hurdle is getting a user to open a file, which is routine phishing.

Which deployments are exposed

Product Status
Microsoft 365 Apps for Enterprise (Click-to-Run) Affected until updated to the November build for its channel
Office LTSC 2021 Affected below 16.0.14334.21048
Office LTSC 2024 Affected below 16.0.17932.21140
Office 2016 / 2019 Not listed by Microsoft. Both are out of support, so absence from the list isn’t evidence that they’re safe
Outlook on the web, mobile, Mac Not listed in Microsoft’s advisory

Prerequisites: the user opens the crafted file in an affected Outlook build, and the workstation can reach the attacker’s server over a protocol that triggers automatic authentication. That second point is where your network controls earn their keep.

Identifying affected systems

Microsoft 365 Apps builds for the 10 November 2026 security release, from Microsoft’s Microsoft 365 Apps security updates page:

Channel Version Build
Current Channel 2610 20527.20196
Monthly Enterprise Channel 2610 / 2609 / 2608 20527.20194 / 20430.20264 / 20326.20320
Semi-Annual Enterprise Channel 2610 / 2609 / 2608 20527.20192 / 20430.20262 / 20326.20318

Builds are full versions like 16.0.20527.20196. On a Windows endpoint you manage:

# Click-to-Run version and the channel it updates from
Get-ItemProperty 'HKLM:\SOFTWARE\Microsoft\Office\ClickToRun\Configuration' |
  Select-Object VersionToReport, CDNBaseUrl, UpdateChannel, Platform

# The Outlook binary itself (adjust the path for 32-bit Office on 64-bit Windows)
(Get-Item 'C:\Program Files\Microsoft Office\root\Office16\OUTLOOK.EXE').VersionInfo.ProductVersion

Across the fleet, use Intune or Configuration Manager software inventory, or Defender Vulnerability Management’s per-CVE view, and count devices still below the fixed build. Don’t assume Click-to-Run auto-update has landed. Devices that are rarely rebooted, with Office open for weeks, often sit on an old build with the new one downloaded but not applied.

Then check whether the attack path works from your network:

# From a standard workstation: can it reach SMB on the internet? It shouldn't.
# Use a host you control as the target.
Test-NetConnection -ComputerName your-test-host.example.com -Port 445

Safe validation in an isolated lab

  1. On a lab VM, install an affected Office build, run the inventory checks above and your scanner, then update and confirm both report the fixed build.
  2. Enable the “Network security: Restrict NTLM: Outgoing NTLM traffic to remote servers” policy in Audit mode. Map a share on a non-domain lab server (net use \\labserver\share) and confirm an event appears in Applications and Services Logs > Microsoft > Windows > NTLM > Operational.
  3. Confirm your egress firewall blocks TCP 445 from the lab workstation to an “external” lab segment, and that the block is logged.
  4. Confirm your EDR or Sysmon pipeline records the network connection with the initiating process, so you’d see OUTLOOK.EXE as the source if it ever happened in production.

Detection and threat hunting

Microsoft hasn’t published IOCs for this campaign. The following hunting leads are behavioral, so no single hit proves exploitation:

  • Outlook talking SMB. Network connections from OUTLOOK.EXE to TCP 445 (or 139) on any address outside your own file servers, especially external IPs. Sysmon Event ID 3 or your EDR’s network telemetry gives you the initiating process.
  • NTLM to the internet. Outgoing NTLM authentication to non-corporate hosts in the NTLM Operational log (with outbound auditing enabled), and blocked or allowed 445 egress in firewall logs.
  • WebDAV fallback. When SMB is blocked, Windows can try WebDAV over HTTP(S) to the same UNC target if the WebClient service is running. Watch for WebClient activity and rundll32.exe ... davclnt.dll processes after a user opens an attachment.
  • Timing. A file attachment opened from Outlook, followed within seconds by an outbound SMB/WebDAV attempt from the same host.
  • Use of the credential. For users who triggered the above: sign-ins from unusual hosts or IPs, NTLM logons (Event ID 4624 with Authentication Package: NTLM) to servers they don’t normally use, and password-spray-like patterns against that one identity.

Remediation

  1. Update Office to the build for each channel above, or to LTSC 2021 16.0.14334.21048 / LTSC 2024 16.0.17932.21140 or later. Verify with inventory rather than trusting the update policy.
  2. If you find signs of a leaked hash, reset that user’s password (that invalidates the captured hash for offline cracking), revoke sessions, and review authentication logs for relay activity around the time of the event.

Interim and long-term hardening

These help against CVE-2026-124194 and the whole class of NTLM-leak bugs:

  • Block outbound SMB (TCP 445) to the internet at the perimeter and on the host firewall. Microsoft has recommended this for years, and it kills the most common path for this class of bug.
  • Restrict outgoing NTLM. Start the “Restrict NTLM: Outgoing NTLM traffic to remote servers” policy in Audit mode, review the results, then move to Deny with an exception list for the servers that genuinely need it.
  • Put privileged accounts in the Protected Users group, which prevents NTLM authentication for its members.
  • Make relay harder: require SMB signing, enable LDAP signing and channel binding, and turn on EPA for AD CS and other web services that accept NTLM.
  • Disable the WebClient service on workstations that don’t need WebDAV.
  • Review mail-flow rules for attached message and Office files from external senders, and remind users that an “email inside an email” from an unknown sender deserves the same suspicion as any other attachment.

This is the second Microsoft collaboration-stack bug this month where the label undersold the risk. The first was SharePoint’s CVE-2026-65660, initially published as spoofing.

References

Verify every version against the vendor advisory before acting. Published 12 Nov 2026.